Skip to content

Daily capacity budget for Sioul — literature and prior art

Correction (5 October 2026, from wellbeing-gain.md). Wu et al. 2017 (experience sampling in daily life) found that people with depression predict pleasure about as they then feel it, both low; what they over-predict is displeasure. The gain half of criterion 7 below (gain under-predicted in low mood) is withdrawn; the anxiety half stands. Plan from ratings given after the activity, and never nudge gain forecasts upward.

Research question: how can a planner measure and calibrate a person's daily capacity for demanding work from self-rated costs per task and per event (cognitive, emotional, anxiety; 0–10), and plan future days within that capacity?

Compiled 2026-10-05. Sections: 1 rating task demands; 2 energy envelope and pacing; 3 learning capacity from history; 4 fatigue, effort and recovery; 5 apps; 6 well-being gain (the floor to fill); 7 recovery pace, time estimation, what to show; then design criteria and open questions. Numbers in criteria point to entries (e.g., 3.16). Verification: "verified" = checked against Crossref or PubMed metadata (title, authors, year, venue) on 2026-10-05. "unverified" = could not confirm; do not cite without checking. Strength scale used below: strong (meta-analysis / large preregistered or multi-site / guideline with systematic review), moderate (several consistent studies, or one large well-run study), weak (one small study, cross-sectional only, case reports), opinion (essay, consensus without data, vendor documentation).

1. Measuring task demands by self-report

1.1 Hart SG, Staveland LE. 1988. Development of NASA-TLX (Task Load Index): results of empirical and theoretical research. Advances in Psychology 52:139–183. doi:10.1016/S0166-4115(08)62386-9 — verified (Crossref).

  • Six subscales rated after a task: mental demand, physical demand, temporal demand, own performance, effort, frustration; originally weighted by 15 pairwise comparisons per rater.
  • Development programme (many lab and simulator studies). Strong as the reference instrument for perceived workload.

1.2 Hart SG. 2006. NASA-Task Load Index (NASA-TLX); 20 years later. Proc Hum Factors Ergon Soc Annu Meet 50(9):904–908. doi:10.1177/154193120605000909 — verified (Crossref).

  • Survey of 550 studies: the TLX is "reasonably easy to use and reliably sensitive to experimentally important manipulations". Many users drop the weights ("Raw TLX", plain mean of subscales); comparisons of raw vs weighted are mixed (more, equally or less sensitive). Origin of the raw-vs-weighted debate: Byers JC, Bittner AC, Hill SG 1989, "Traditional and raw task load index (TLX) correlations: are paired comparisons necessary?" (Advances in Industrial Ergonomics and Safety I, book chapter; unverified in registries, widely cited).
  • Review. Moderate. Relevance: per-person weighting of subscales adds burden for little gain; an unweighted combination is the accepted default.

1.3 Galy E, Paxion J, Berthelon C. 2018. Measuring mental workload with the NASA-TLX needs to examine each dimension rather than relying on the global score: an example with driving. Ergonomics 61(4):517–527. doi:10.1080/00140139.2017.1369583 — verified (PubMed 28817353).

  • Situation complexity, experience, tension and alertness moved different TLX subscales in different directions; the authors question the use of the global score.
  • One driving-simulator study. Weak-to-moderate. Supports keeping Sioul's three scales as three budgets rather than collapsing them.

1.4 Grier RA. 2015. How high is high? A meta-analysis of NASA-TLX global workload scores. Proc Hum Factors Ergon Soc Annu Meet 59(1):1727–1731. doi:10.1177/1541931215591373 — verified (Crossref).

  • Distribution of >1000 global TLX scores from >200 publications, by task type, so a score can be read against what others report.
  • Descriptive meta-analysis. Moderate. Relevance: norms exist for workload scores; none exist for Sioul's 0–10 per-task scales, so the scales must be read relative to the person's own history.

1.5 Paas FGWC. 1992. Training strategies for attaining transfer of problem-solving skill in statistics: a cognitive-load approach. J Educ Psychol 84(4):429–434. doi:10.1037/0022-0663.84.4.429 — verified. Paas F, van Merriënboer JJG, Adam JJ. 1994. Measurement of cognitive load in instructional research. Percept Mot Skills 79(1):419–430. doi:10.2466/pms.1994.79.1.419 — verified. Paas F, Tuovinen JE, Tabbers H, Van Gerven PWM. 2003. Cognitive load measurement as a means to advance cognitive load theory. Educ Psychol 38(1):63–71. doi:10.1207/S15326985EP3801_8 — verified.

  • One item, 9 points ("very, very low" to "very, very high mental effort"), asked right after a task: reliability .90 (1992) and α = .82 (1994), sensitive to small differences in load; the 2003 review finds such single-item ratings reliable, non-intrusive and the most used load measure in its field. (Reliability figures as reported by secondary sources and the 2003 review.)
  • Lab studies with students, then many replications. Moderate-to-strong for "a single 0–10-like effort rating per task is usable".

1.6 van Gog T, Kirschner F, Kester L, Paas F. 2012. Timing and frequency of mental effort measurement: evidence in favour of repeated measures. Appl Cogn Psychol 26(6):833–839. doi:10.1002/acp.2883 — verified (Crossref).

  • Four experiments: one rating given after a series of tasks is higher than the mean of ratings given after each task; knowing in advance that one will rate at the end lowers scores. "Repeatedly measuring mental effort after each task in the series seems to be preferable."
  • Experimental, student samples. Moderate.

1.7 Schmeck A, Opfermann M, van Gog T, Paas F, Leutner D. 2015. Measuring cognitive load with subjective rating scales during problem solving: differences between immediate and delayed ratings. Instr Sci 43(1):93–114. doi:10.1007/s11251-014-9328-3 — verified (Crossref).

  • Two experiments (n = 168, 107): delayed ratings of effort and difficulty were higher than the mean of the six immediate ratings and were best predicted by the most complex problems.
  • Experimental. Moderate. With 1.6: a day's felt load is not the average of its tasks; it is pulled up by the hardest ones.

1.8 Fredrickson BL, Kahneman D. 1993. Duration neglect in retrospective evaluations of affective episodes. J Pers Soc Psychol 65(1):45–55. doi:10.1037/0022-3514.65.1.45 — verified (PubMed 8355141). Redelmeier DA, Kahneman D. 1996. Patients' memories of painful medical treatments. Pain 66(1):3–8. doi:10.1016/0304-3959(96)02994-6 — verified (PubMed 8857625).

  • Remembered unpleasantness tracks the peak and the end of an episode; duration barely counts ("as if duration did not matter"; colonoscopy n = 154, lithotripsy n = 133).
  • Experiments and clinical cohort. Strong (much replicated). Relevance: a rating given after a task describes its intensity, not its total cost; the app must multiply by duration itself (as session-RPE does, 3.2) instead of asking "how costly was this task overall".

1.9 de Jonge J, Dormann C. 2006. Stressors, resources, and strain at work: a longitudinal test of the triple-match principle. J Appl Psychol 91(6):1359–1374. doi:10.1037/0021-9010.91.5.1359 — verified (PubMed 17100490).

  • DISC model: demands, resources and strain each come in cognitive, emotional and physical kinds. Resources buffered demands in 33.3% of tests when all three matched, 16.7% with a double match, 0% with no match; most consistent for emotional and physical matches.
  • Two longitudinal surveys, n = 280 and 267 health-care workers. Moderate. Best published support for splitting demands by kind, and for "recover in the domain you spent" (an emotionally heavy task calls for an emotionally restoring activity, not just any pause).

1.10 Kristensen TS, Hannerz H, Høgh A, Borg V. 2005. The Copenhagen Psychosocial Questionnaire — a tool for the assessment and improvement of the psychosocial work environment. Scand J Work Environ Health 31(6):438–449. doi:10.5271/sjweh.948 — verified (PubMed 16425585).

  • Built from a representative sample of 1858 Danish employees; "a novel feature of the COPSOQ is the development of five different scales on demands at work" (quantitative, cognitive, emotional, demands for hiding emotions, sensorial), with good reliability and "very little overlap between the scales".
  • Large validation survey. Moderate-to-strong as established practice: occupational health already measures cognitive and emotional demands separately, and treats hiding emotions as its own demand.

1.11 Wickens CD. 2008. Multiple resources and mental workload. Hum Factors 50(3):449–455. doi:10.1518/001872008X288394 — verified (Crossref).

  • Attention draws on several partly separate resource pools; tasks interfere most when they compete for the same pool. The alternative, a single general effort pool (Kahneman 1973, Attention and Effort), explains the general part of fatigue.
  • Theory with decades of dual-task evidence. Moderate. Relevance: both views hold in part, so check each scale against its own budget and the total against a total budget.

1.12 Richards JM, Gross JJ. 2000. Emotion regulation and memory: the cognitive costs of keeping one's cool. J Pers Soc Psychol 79(3):410–424. doi:10.1037/0022-3514.79.3.410 — verified (Crossref). Eysenck MW, Derakshan N, Santos R, Calvo MG. 2007. Anxiety and cognitive performance: attentional control theory. Emotion 7(2):336–353. doi:10.1037/1528-3542.7.2.336 — verified (Crossref).

  • Suppressing emotion impairs memory for what happens meanwhile; anxiety impairs attentional control, costing efficiency (more effort for the same result) before effectiveness.
  • Experiments; theory with broad support. Moderate. Relevance: the three scales are correlated by mechanism; emotional and anxiety costs also draw on cognitive capacity. Expect positive correlations between scales and do not treat them as independent budgets only.

1.13 Grandey AA. 2000. Emotional regulation in the workplace: a new way to conceptualize emotional labor. J Occup Health Psychol 5(1):95–110. doi:10.1037/1076-8998.5.1.95 — verified. Hülsheger UR, Schewe AF. 2011. On the costs and benefits of emotional labor: a meta-analysis of three decades of research. J Occup Health Psychol 16(3):361–389. doi:10.1037/a0022876 — verified (PubMed 21728441).

  • Emotional labour = regulating felt and shown emotion to meet expectations. Meta-analysis (95 studies, 494 correlations): emotion-rule dissonance and surface acting (showing what one does not feel) relate strongly to impaired well-being (ρ .39–.48); deep acting only weakly.
  • Meta-analysis of mostly cross-sectional studies. Strong for association, weak for causation. Relevance: the anchor for "emotional cost" should name masking/pretending (the costly part), not only sad or tense content. Masking is also named as a cause in autistic burnout (2.17).

1.14 Steel P. 2007. The nature of procrastination: a meta-analytic and theoretical review of quintessential self-regulatory failure. Psychol Bull 133(1):65–94. doi:10.1037/0033-2909.133.1.65 — verified (PubMed 17201571).

  • 691 correlations: "strong and consistent predictors of procrastination were task aversiveness, task delay, self-efficacy, and impulsiveness".
  • Meta-analysis. Strong. Relevance: a high anxiety/aversiveness rating predicts that a task will be postponed, independent of its cognitive cost.

1.15 Blunt AK, Pychyl TA. 2000. Task aversiveness and procrastination: a multi-dimensional approach to task aversiveness across stages of personal projects. Pers Individ Dif 28(1):153–167. doi:10.1016/S0191-8869(99)00091-4 — verified (Crossref). Sirois F, Pychyl T. 2013. Procrastination and the priority of short-term mood regulation: consequences for future self. Soc Personal Psychol Compass 7(2):115–127. doi:10.1111/spc3.12011 — verified (Crossref).

  • Boring, frustrating and resented projects are aversive at every stage; at the planning stage lack of personal meaning matters, at the doing stage lack of structure/clarity matters (summary via secondary sources; abstract not in registries). Procrastination is short-term mood repair at the expense of the future self.
  • Personal-projects survey (students) and theory review. Moderate. Relevance: anxiety/aversiveness can be lowered by giving a task a clear first step; the planner can treat "unclear" as a cost.

1.16 Rachman S. 1994. The overprediction of fear: a review. Behav Res Ther 32(7):683–690. doi:10.1016/0005-7967(94)90025-6 — verified (PubMed 7980354). LaFreniere LS, Newman MG. 2020. Exposing worry's deceit: percentage of untrue worries in generalized anxiety disorder treatment. Behav Ther 51(3):413–423. doi:10.1016/j.beth.2019.07.003 — verified (PubMed 32402257).

  • People commonly predict more fear than they then feel, and predictions fall with repeated exposure. In GAD (n = 29, EMA over 30 days) 91.4% of worries did not come true, and seeing this predicted improvement.
  • Review; one small EMA study. Moderate for overprediction, weak for the 91% figure. Relevance: an anxiety rating given when the task is created will usually exceed what is felt; asking again after the task and showing the person the difference is a known therapeutic move, while the load computation should use the after-task rating.

1.17 Ratings on 0–10 scales: reliability and meaningful change. Hjermstad MJ, Fayers PM, Haugen DF, et al. 2011. Studies comparing numerical rating scales, verbal rating scales, and visual analogue scales for assessment of pain intensity in adults: a systematic literature review. J Pain Symptom Manage 41(6):1073–1093. doi:10.1016/j.jpainsymman.2010.08.016 — verified (PubMed 21621130). Farrar JT, Young JP, LaMoreaux L, Werth JL, Poole MR. 2001. Clinical importance of changes in chronic pain intensity measured on an 11-point numerical pain rating scale. Pain 94(2):149–158. doi:10.1016/S0304-3959(01)00349-9 — verified (PubMed 11690728). Mattera E, Zaboski B. 2025. Rethinking the Subjective Units of Distress Scale: validity and clinical utility of the SUDS. Clin Pract 15(7):123. doi:10.3390/clinpract15070123 — verified (PubMed 40710033).

  • 0–10 numeric scales have better compliance and responsiveness than visual or verbal scales (54 studies). A change that patients call important is ~2 points or ~30% (2724 patients, 10 trials). The SUDS (the 0–100/0–10 anxiety scale used to pace exposure therapy) is ubiquitous but has weak psychometrics: poorly defined occasions, construct under-representation.
  • Systematic review; large pooled analysis; critical review. Strong / strong / opinion-level critique. Relevance: treat a 1-point difference as noise; do not react to, or display, changes under ~2 points or ~30%; anchor each scale with words at 0, 5 and 10 and say when to rate (right after the task).

1.18 How to combine the three scales — what the sources allow.

  • No study validates a cognitive + emotional + anxiety composite for planning. The nearest evidence: unweighted means are the accepted default when several workload subscales are combined (1.2); subscales carry different information (1.3, 1.9, 1.10); resources are partly shared (1.11, 1.12).
  • Inference (mine): keep three budgets and check each; also learn a separate total budget from the unweighted sum of the scales on good days and check the day's sum against it (shared pool: a day cannot fill all three budgets at once unless the history shows it can); use the maximum of the three normalised loads to rank tasks heavy vs light (a task is as heavy as its worst dimension, and remembered load follows peaks: 1.7, 1.8). Per-person weights only after months of data, if ever.

2. Energy envelope, pacing, and the apps built on them

2.1 Miserandino C. 2003. The spoon theory. But You Don't Look Sick (essay). https://www.butyoudontlooksick.com/articles/written-by-christine/the-spoon-theory/ — exists (URL resolves); not peer-reviewed.

  • A person with lupus explains a day as a fixed handful of spoons; each act costs spoons; borrowing from tomorrow leaves tomorrow short.
  • Opinion/metaphor. No peer-reviewed study of "spoons" as a measure was found in PubMed (search "spoon theory"[tiab], 2026-10-05). Its value for Sioul is language, not calibration: it names a daily budget, carry-over to the next day, and the fact that healthy people do not count.

2.2 Jason LA, Brown M, Brown A, et al. 2013. Energy conservation/envelope theory interventions to help patients with myalgic encephalomyelitis/chronic fatigue syndrome. Fatigue: Biomed Health Behav 1(1–2):27–42. doi:10.1080/21641846.2012.733602 — verified (PubMed 23504301).

  • Fifteen years of work on helping people "self-monitor and self-regulate energy expenditures and learn to pace activities and stay within their energy envelope"; reviews trials and case series in which staying inside the envelope went with better quality of life.
  • Narrative review by the theory's authors. Weak-to-moderate (small trials, risk of allegiance bias).

2.3 O'Connor K, Sunnquist M, Nicholson L, Jason LA, Newton JL, Strand EB. 2019. Energy envelope maintenance among patients with myalgic encephalomyelitis and chronic fatigue syndrome: implications of limited energy reserves. Chronic Illn 15(1):51–60. doi:10.1177/1742395317746470 — verified (PubMed 29231037; full text read).

  • Operational definition used in this literature: two self-ratings for the past week, available energy and expended energy, each 1–100 where 100 = pre-illness level (test–retest .81 and .64). Energy quotient = expended / available × 100; >100 = overextended, <100 = underextended. In n = 434 (four samples), people with more available energy functioned better, but much less so if they overextended.
  • Cross-sectional. Weak for causation. Two lessons for Sioul: (a) the field asks the person directly for "available" energy rather than inferring it from load; (b) the "expended" rating is the less reliable of the two (.64).

2.4 Brown M, Khorana N, Jason LA. 2011. The role of changes in activity as a function of perceived available and expended energy in nonpharmacological treatment outcomes for ME/CFS. J Clin Psychol 67(3):253–260. doi:10.1002/jclp.20744 — verified (PubMed 21254053). Jason LA, Tryon WW, Taylor RR, et al. 1999. Monitoring and assessing symptoms of chronic fatigue syndrome: use of time series regression. Psychol Rep 85(1):121–130. doi:10.2466/pr0.1999.85.1.121 — verified (PubMed 10575979).

  • n = 44: people inside their envelope before treatment improved more in physical function and fatigue. Single-case daily/hourly series (16 months, with actigraphy): perceived energy, physical exertion and mental exertion each predicted fatigue within and between days; as the person brought expenditure down to available energy, fatigue fell and available energy rose.
  • Small cohort; single case. Weak. But it is the closest published precedent to Sioul's design: daily self-rated mental and physical exertion against a self-rated envelope, analysed as a time series.

2.5 Goudsmit EM, Nijs J, Jason LA, Wallman KE. 2012. Pacing as a strategy to improve energy management in myalgic encephalomyelitis/chronic fatigue syndrome: a consensus document. Disabil Rehabil 34(13):1140–1147. doi:10.3109/09638288.2011.635746 — verified (PubMed 22181560).

  • Pacing (adapting activity to the energy available, with rest before exhaustion) is "consistently rated as one of the most helpful" strategies in patient surveys; proposed to stabilise the illness and avoid post-exertional malaise.
  • Consensus/narrative review. Opinion-level, by researchers aligned with pacing.

2.6 NICE. 2021. Myalgic encephalomyelitis (or encephalopathy)/chronic fatigue syndrome: diagnosis and management. NICE guideline NG206. https://www.nice.org.uk/guidance/ng206 — verified (recommendations page read 2026-10-05).

  • Energy management "includes all types of activity (cognitive, physical, emotional and social)"; "each person has a different and fluctuating energy limit and they are experts in judging their own limits"; uses "a flexible, tailored approach so that activity is never automatically increased but is maintained or adjusted (upwards after a period of stability or downwards when symptoms are worse)"; "it can take weeks, months or sometimes even years to reach stabilisation or to increase tolerance" (1.11.2).
  • 1.11.4: "agree a sustainable level of activity as the first step, which may mean reducing activity; plan periods of rest and activity, and incorporate the need for pre-emptive rest; alternate and vary between different types of activity and break activities into small chunks." 1.11.7: make self-monitoring easy (tracker, heart-rate monitor or diary). 1.11.13: set the baseline "at a level that does not worsen their symptoms", initially go below it, maintain "successfully for a period of time before attempting to increase it". 1.11.14: do not offer programmes "that use fixed incremental increases" (graded exercise therapy). 1.11.15: after a flare-up, reduce to within current limits and set a new baseline only once symptoms stabilise. PEM "is often delayed in onset by hours or days … has a prolonged recovery time that may last hours, days, weeks or longer" (box 2). Relapse "can lead to a long-term reduction in the person's energy limits" (glossary).
  • National guideline from systematic reviews; the committee rated the trial evidence for all non-drug treatments low to very low quality, so much of it rests on expert and patient testimony. Strong as practice standard, weak-to-moderate as evidence. These sentences translate almost one-to-one into planner rules.

2.7 White PD, Goldsmith KA, Johnson AL, et al. 2011. Comparison of adaptive pacing therapy, cognitive behaviour therapy, graded exercise therapy, and specialist medical care for chronic fatigue syndrome (PACE): a randomised trial. Lancet 377(9768):823–836. doi:10.1016/S0140-6736(11)60096-2 — verified (PubMed 21334061). Critique: Wilshire CE, Kindlon T, Courtney R, et al. 2018. Rethinking the treatment of chronic fatigue syndrome — a reanalysis and evaluation of findings from a recent major trial of graded exercise and CBT. BMC Psychol 6:6. doi:10.1186/s40359-018-0218-3 — verified (PubMed 29562932). Also Vink M, Vink-Niese A. 2018. Graded exercise therapy for ME/CFS is not effective and unsafe. Re-analysis of a Cochrane review. Health Psychol Open 5(2). doi:10.1177/2055102918805187 — verified (Crossref).

  • PACE (n = 641): GET and CBT lowered fatigue by ~3 points and raised physical function by 7–9 points vs medical care alone at 52 weeks; adaptive pacing therapy did not differ from medical care. Reanalysis using the protocol's own primary outcome: CBT and GET no longer beat control after correction for multiple comparisons, recovery rates were low and similar, effects were confined to self-report and gone by two years.
  • One large RCT, contested; reanalyses by critics (some patient-advocates). Net: neither "increase on a schedule" (GET) nor PACE's version of pacing has strong evidence; NICE 2021 then advised against fixed increments. For Sioul: the budget should rise only when the person's own record shows stable days, never on a timetable.

2.8 Chu L, Valencia IJ, Garvert DW, Montoya JG. 2018. Deconstructing post-exertional malaise in myalgic encephalomyelitis/chronic fatigue syndrome: a patient-centered, cross-sectional survey. PLoS ONE 13(6):e0197811. doi:10.1371/journal.pone.0197811 — verified (PubMed 29856774).

  • n = 150: 90% had PEM after physical and cognitive exertion and after emotional distress; onset and duration varied; 11% had a consistent delay of ≥24 h before onset; 84% had PEM lasting ≥24 h.
  • Cross-sectional survey, one clinic. Moderate for the description. Relevance: cognitive and emotional exertion trigger crashes too (supports Sioul's scales), and a bad day may be the bill for one or two days earlier, so the estimator must look back 1–3 days when a bad day is reported.

2.9 Parker M, Sawant HB, Flannery T, et al. 2023. Effect of using a structured pacing protocol on post-exertional symptom exacerbation and health status in a longitudinal cohort with the post-COVID-19 syndrome. J Med Virol 95(1):e28373. doi:10.1002/jmv.28373 — verified (PubMed 36461167; full text read).

  • WHO Borg CR-10 five-phase pacing protocol: each phase caps perceived exertion; "stay at each phase for a minimum of 7 days"; if exertion exceeds the phase's upper limit, stop that activity. In 31 people with long COVID (mean 17 months), crashes fell from 3.4/week to 1.1/week over 6 weeks (−16%/week), across physical, cognitive and emotional triggers; EQ-5D 51.4 → 60.6.
  • Uncontrolled cohort, n = 31; authors wrote the WHO booklet (declared). Weak. Practice precedent for "a minimum period of stability before stepping up" (7 days) and for capping intensity, not only volume.

2.10 Décary S, Gaboury I, Poirier S, et al. 2021. Humility and acceptance: working within our limits with long COVID and myalgic encephalomyelitis/chronic fatigue syndrome. J Orthop Sports Phys Ther 51(5):197–200. doi:10.2519/jospt.2021.0106 — verified (Crossref).

  • Physiotherapists' viewpoint: exertion intolerance and PEM call for working within limits, not progressive overload.
  • Opinion. Shows the sports-rehab profession itself warns against transposing training-load progression to post-viral fatigue.

2.11 Abonie US, Sandercock GRH, Heesterbeek M, Hettinga FJ. 2020. Effects of activity pacing in patients with chronic conditions associated with fatigue complaints: a meta-analysis. Disabil Rehabil 42(5):613–622. doi:10.1080/09638288.2018.1504994 — verified (PubMed 30449204).

  • 6 studies: medium effect on fatigue after treatment (SMD 0.50), marginal at follow-up (0.34); no meaningful effect on physical function or activity (0.08–0.30); minimal contact worked as well as intensive.
  • Small meta-analysis. Moderate-to-weak. Pacing reduces fatigue modestly; it does not by itself raise capacity.

2.12 Kos D, van Eupen I, Meirte J, et al. 2015. Activity pacing self-management in chronic fatigue syndrome: a randomized controlled trial. Am J Occup Ther 69(5):6905290020. doi:10.5014/ajot.2015.016287 — verified (PubMed 26356665).

  • 33 women, pacing self-management vs relaxation: satisfaction with performance of chosen activities improved more with pacing (effect size 0.74).
  • Small RCT. Weak.

2.13 Andrews NE, Strong J, Meredith PJ. 2012. Activity pacing, avoidance, endurance, and associations with patient functioning in chronic pain: a systematic review and meta-analysis. Arch Phys Med Rehabil 93(11):2109–2121. doi:10.1016/j.apmr.2012.05.029 — verified (PubMed 22728699). Murphy SL, Kratz AL. 2014. Activity pacing in daily life: a within-day analysis. Pain 155(12):2630–2637. doi:10.1016/j.pain.2014.09.028 — verified (PubMed 25267209).

  • 41 studies: both avoidance and overactivity (persisting until symptoms flare — the "boom" of boom–bust) go with worse outcomes; pacing goes with better psychological functioning but more pain and disability. Daily-life study (n = 162, 5×/day for 5 days): people paced when symptoms were already high, and this naturalistic pacing did not lower later symptoms.
  • Meta-analysis of correlations; EMA study. Moderate. Relevance: pacing that only reacts to symptoms (stopping when already depleted) does not work; it has to be planned ahead (pre-emptive rest, NICE 1.11.4). That is the planner's job.

2.14 Antcliff D, Keenan A-M, Keeley P, Woby S, McGowan L. 2021. Testing a newly developed activity pacing framework for chronic pain/fatigue: a feasibility study. BMJ Open 11(12):e045398. doi:10.1136/bmjopen-2020-045398 — verified (PubMed 34880007). Framework built in Antcliff et al. 2019, Musculoskeletal Care 17(4):335–345 and 354–362, doi:10.1002/msc.1421 and doi:10.1002/msc.1430 — verified (Crossref).

  • Standardised pacing instructions across professions (survey + nominal group); single-arm feasibility (107 recruited, 39% attrition): pacing and all symptoms improved by end of programme, smaller gains at 3 months.
  • Uncontrolled. Weak. Shows pacing is now being standardised, still without RCT proof.

2.15 Energy conservation trials outside ME/CFS. Mathiowetz VG, Finlayson ML, Matuska KM, Chen HY, Luo P. 2005. Randomized controlled trial of an energy conservation course for persons with multiple sclerosis. Mult Scler 11(5):592–601. doi:10.1191/1352458505ms1198oa — verified (PubMed 16193899). Barsevick AM, Dudley W, Beck S, et al. 2004. A randomized clinical trial of energy conservation for patients with cancer-related fatigue. Cancer 100(6):1302–1310. doi:10.1002/cncr.20111 — verified (PubMed 15022300).

  • MS (n = 169, crossover RCT): a 6-week course (prioritise, plan, rest, pace) reduced fatigue impact and raised vitality and self-efficacy. Cancer (n = 396): energy conservation and activity management gave "a modest but significant" drop in fatigue, no change in functional performance.
  • RCTs. Moderate. Planning activity against a limited budget helps fatigue a little; it does not restore function by itself.

2.16 Kluger BM, Krupp LB, Enoka RM. 2013. Fatigue and fatigability in neurologic illnesses: proposal for a unified taxonomy. Neurology 80(4):409–416. doi:10.1212/WNL.0b013e31827f07be — verified (PubMed 23339207).

  • Separates perceived fatigue (a feeling) from fatigability (how fast performance or perception changes with a given amount of activity).
  • Expert taxonomy. Opinion. Relevance: Sioul's "capacity" is a fatigability estimate (load tolerated before symptoms), which is why it must be inferred from load and outcome together, not from how tired the person says they feel in the morning alone.

2.17 Autistic burnout. Raymaker DM, Teo AR, Steckler NA, et al. 2020. "Having all of your internal resources exhausted beyond measure and being left with no clean-up crew": defining autistic burnout. Autism Adulthood 2(2):132–143. doi:10.1089/aut.2019.0079 — verified (PubMed 32851204). Higgins JM, Arnold SR, Weise J, Pellicano E, Trollor JN. 2021. Defining autistic burnout through experts by lived experience: grounded Delphi method investigating #AutisticBurnout. Autism 25(8):2356–2369. doi:10.1177/13623613211019858 — verified (PubMed 34088219). Mantzalas J, Richdale AL, Dissanayake C. 2022. A conceptual model of risk and protective factors for autistic burnout. Autism Res 15(6):976–987. doi:10.1002/aur.2722 — verified (Crossref).

  • Raymaker (19 interviews + 19 internet sources, community-based participatory research): chronic exhaustion, loss of skills, reduced tolerance to stimuli, caused by "life stressors that added to the cumulative load" with no relief, until "expectations outweigh abilities". Recovery came with acceptance, "time off/reduced expectations", and unmasking. Higgins (Delphi): exhaustion, withdrawal, cognitive problems, loss of daily living skills; distinct from depression, and depression treatments may worsen it.
  • Qualitative/Delphi. Weak-to-moderate (well-done qualitative). Relevance: cumulative load, sensory load and masking are costs; "reduced expectations" is the treatment, so a planner must be able to plan days well below past capacity without signalling failure.

2.18 Energy accounting (autism practice). Toudal M, Attwood T. 2024. Energy Accounting: Stress Management and Mental Health Monitoring for Autism and Related Conditions. London: Jessica Kingsley. ISBN 9781787757752 — verified (publisher listing).

  • Activities are logged as withdrawals (draining) and deposits (restoring), with the aim of keeping a positive daily balance.
  • Clinical method book; no outcome trial found. Opinion. Direct precedent for Sioul's cost/gain pairing (section 6).

2.19 Visible (pacing app for ME/CFS and long COVID). Vendor pages consulted 2026-10-05: https://www.makevisible.com/blog/introducing-the-morning-stability-score ; https://help.makevisible.com/en/articles/10632758-can-i-use-visible-if-i-have-pots

  • Morning Stability Score (0–5) from changes in HRV, resting heart rate, sleep score and evening symptom trends, each compared with "your previous 14 days of data … a rolling time window"; they moved from 7 to 14 days because the 7-day window "is significantly impacted by multi-day crashes". With an armband, daytime heart rate feeds a daily "PacePoints" budget; zones and budget are set from defaults and, if they feel wrong, adjusted by support staff after the first 4 days of data (not automatically). User reports (S4ME forum, unverified) say points accrue per minute above an activity threshold and ×4 above an over-exertion threshold.
  • Vendor documentation. Opinion. Two takeaways: a 14-day baseline was chosen over 7 days for robustness to crashes; even the leading pacing app keeps the budget mostly hand-set rather than learned.

2.20 Bearable (symptom and mood tracker). Vendor support pages consulted 2026-10-05: https://bearable.app/support/howto/configure-and-enter-data-into-bearable/ (and linked support articles).

  • Energy rated daily 1–5; "insights" compare scores on days with vs without a factor and need at least 3 days with and 3 without; 30 days recommended.
  • Vendor documentation. Opinion. It correlates, it does not budget or plan. No published outcome study found.

2.21 Patient-organisation practice: start at half, rest before known exertion. Seltzer J (#MEAction). Pacing and management guide for ME/CFS. Distributed by Unity Health Toronto, Outpatient Post-COVID Condition Rehabilitation Program (PDF created 2021, revised 2023). https://unityhealth.to/wp-content/uploads/2023/06/pacing-and-management-guide-for-ME-CFS.pdf — exists (downloaded and read 2026-10-05).

  • "Decreasing activity by half to start with can help the person with ME find a baseline — with the understanding that the baseline can become better or worse over time." PEM is triggered by physical, cognitive, sensory, emotional, orthostatic and environmental load and "may be delayed by approximately 24 hours". "Radical rest": being inactive before planned exertions (a wedding, a talk) and planning low-energy days after them.
  • Patient-led practice guide. Opinion. With NICE 1.11.4 ("pre-emptive rest") it gives two planner rules: start the budget low (about half of current boom-day activity) and lighten the days around an unavoidable heavy event. It also lists sensory and orthostatic load, which Sioul's three scales do not cover.

3. Learning capacity from history (sports load monitoring and its critiques)

Each entry: reference — finding — study type and size — strength.

3.1 Foster C. 1998. Monitoring training in athletes with reference to overtraining syndrome. Med Sci Sports Exerc 30(7):1164–1168. doi:10.1097/00005768-199807000-00023 — verified (PubMed 9662690).

  • Load = session RPE × duration. Defines monotony = daily mean load / SD of daily load over a week, and strain = weekly load × monotony. "A high percentage of illnesses could be accounted for when individual athletes exceeded individually identifiable training thresholds, mostly related to the strain of training."
  • Observational, n = 25 athletes. Weak, but the monotony idea (same load every day is worse than alternating hard and easy) is the sports root of "heavy then light".

3.2 Foster C, Florhaug JA, Franklin J, et al. 2001. A new approach to monitoring exercise training. J Strength Cond Res 15(1):109–115. PubMed 11708692 — verified.

  • Session-RPE (Borg CR-10 rating of the whole session, asked ~30 min after) × minutes tracks a heart-rate-based load "highly consistently" across steady, interval and team-sport exercise; absolute sRPE scores run higher than HR scores.
  • Small validation study (two parts, cycling and basketball). Moderate when combined with 3.3.

3.3 Haddad M, Stylianides G, Djaoui L, Dellal A, Chamari K. 2017. Session-RPE method for training load monitoring: validity, ecological usefulness, and influencing factors. Front Neurosci 11:612. doi:10.3389/fnins.2017.00612 — verified (PubMed 29163016).

  • Narrative review of 36 validation studies: sRPE × duration shows good validity, reliability and internal consistency across sports, ages and levels; can "stand alone" as a load measure.
  • Review (not systematic meta-analysis). Moderate. Relevance: a single 0–10 rating × duration is the most validated self-report load unit in any field. Sioul's per-task 0–10 rating × tracked time is the same construction.

3.4 Banister model (fitness–fatigue, impulse–response). Calvert TW, Banister EW, Savage MV, Bach T. 1976. A systems model of the effects of training on physical performance. IEEE Trans Syst Man Cybern SMC-6(2):94–102. doi:10.1109/TSMC.1976.5409179 — verified. (The 1975 Aust J Sports Med 7:57–61 paper has no DOI — unverified by registry, widely cited.)

  • Performance(t) = baseline + k1·Σ load·e^(−Δt/τ1) − k2·Σ load·e^(−Δt/τ2): each day's load adds a slow "fitness" and a fast "fatigue" response, both decaying exponentially.
  • Model, small samples. It is the theoretical origin of exponential kernels in load monitoring.

3.5 Hellard P, Avalos M, Lacoste L, et al. 2006. Assessing the limitations of the Banister model in monitoring training. J Sports Sci 24(5):509–520. doi:10.1080/02640410500244697 — verified (PubMed 16608765).

  • Fit looks good (R² = 0.79) but parameters are poorly identified: the time constants came with wide 95% CIs, e.g. 38 (17–59), 19 (6–32), 19 (7–35) and 43 (25–61) days; some parameters were so correlated that this made "their interpretation worthless".
  • 9 elite swimmers, one season. Moderate as a warning: with months of daily data per person, a 4-parameter decay model still cannot be pinned down. Sioul should not fit per-person time constants from a few weeks of data.

3.6 Busso T. 2003. Variable dose-response relationship between exercise training and performance. Med Sci Sports Exerc 35(7):1188–1195. doi:10.1249/01.MSS.0000074465.13621.37 — verified (PubMed 12840641).

  • Adding a term where fatigue per unit load grows with recent load improves fit and yields an inverted-U between daily training amount and performance: beyond some daily dose, more load lowers performance.
  • n = 6, 15 weeks. Weak alone, but it is the formal statement that the cost of a unit of load is higher when recent load is high.

3.7 Kolossa D, Bin Azhar MA, Rasche C, et al. 2017. Performance estimation using the fitness-fatigue model with Kalman filter feedback. Int J Comput Sci Sport 16(2):117–129. doi:10.1515/ijcss-2017-0010 — verified (Crossref).

  • The Busso/Banister model is a linear time-variant state-space model, so a Kalman filter can correct the fitness and fatigue estimates whenever a performance measurement arrives; this "clearly" improves tracking and prediction.
  • Methods paper with simulations/data. Moderate as engineering. Relevance: a state-space filter is the principled way to mix load (input) with occasional outcome reports (measurements) and to skip days without data.

3.8 Gabbett TJ. 2016. The training—injury prevention paradox: should athletes be training smarter and harder? Br J Sports Med 50(5):273–280. doi:10.1136/bjsports-2015-095788 — verified (PubMed 26758673).

  • Popularised the acute:chronic workload ratio (ACWR: last 7 days ÷ last 28 days) and its "sweet spot" (about 0.8–1.3; risk rises above ~1.5). Argues that "excessive and rapid increases in training loads" cause many injuries, and that under-training also raises risk.
  • Narrative review / opinion built on cohort studies (e.g., Hulin et al. 2016, Br J Sports Med 50:231–236, doi:10.1136/bjsports-2015-094817 — verified, elite rugby league). Weak-to-moderate; see critiques 3.11–3.13.

3.9 Williams S, West S, Cross MJ, Stokes KA. 2017. Better way to determine the acute:chronic workload ratio? Br J Sports Med 51(3):209–210. doi:10.1136/bjsports-2016-096589 — verified (PubMed 27650255).

  • Letter. Rolling averages weight every day in the window equally and ignore that fitness and fatigue decay; proposes EWMA: EWMA_today = load_today·λ + (1 − λ)·EWMA_yesterday, λ = 2/(N + 1), N = 7 (acute) and 28 (chronic). (Formula as reproduced by secondary sources; full text behind paywall.)
  • Opinion/methods letter. Weak as evidence, standard as practice.

3.10 Murray NB, Gabbett TJ, Townshend AD, Blanch P. 2017. Calculating acute:chronic workload ratios using exponentially weighted moving averages provides a more sensitive indicator of injury likelihood than rolling averages. Br J Sports Med 51(9):749–754. doi:10.1136/bjsports-2016-097152 — verified (PubMed 28003238).

  • 59 Australian footballers, 2 seasons, GPS loads: EWMA-based ratio explained more injury variance than rolling averages; both linked very high spikes (ratio > 2.0) to injury.
  • Single-club cohort. Weak-to-moderate, and inherits the ACWR critiques below.

3.11 Impellizzeri FM, Tenan MS, Kempton T, Novak A, Coutts AJ. 2020. Acute:chronic workload ratio: conceptual issues and fundamental pitfalls. Int J Sports Physiol Perform 15(6):907–913. doi:10.1123/ijspp.2019-0864 — verified (PubMed 32502973).

  • "There is no evidence supporting the use of ACWR in training-load-management systems or for training recommendations aimed at reducing injury risk." No study has estimated a causal effect; the ratio is not properly normalised, "adds noise and creates statistical artifacts".
  • Methodological critique. Strong as a critique of the ratio; it does not show that load spikes are harmless.

3.12 Impellizzeri FM, Woodcock S, Coutts AJ, Fanchini M, McCall A, Vigotsky AD. 2021. What role do chronic workloads play in the acute to chronic workload ratio? Time to dismiss ACWR and its underlying theory. Sports Med 51(3):581–592. doi:10.1007/s40279-020-01378-6 — verified (PubMed 33332011).

  • Dividing acute load by a fixed number or by random "chronic loads" gives injury associations as strong as the real ratio (OR 1.95 vs 2.45); neither ratio nor acute load predicts better than chance (c-statistic 0.574/0.544 vs 0.5). The chronic load adds nothing.
  • Reanalysis of published data. Strong as a critique. Relevance for Sioul: a moving average of past load is a description of habit, not a measure of capacity ("fitness"), unless outcomes are brought in.

3.13 Lolli L, Batterham AM, Hawkins R, et al. 2019. Mathematical coupling causes spurious correlation within the conventional acute-to-chronic workload ratio calculations. Br J Sports Med 53(15):921–922. doi:10.1136/bjsports-2017-098110 — verified (PubMed 29101104). Also Carey DL, Crossley KM, Whiteley R, et al. 2018. Modeling training loads and injuries: the dangers of discretization. Med Sci Sports Exerc 50(11):2267–2276. doi:10.1249/MSS.0000000000001685 — verified (Crossref).

  • When the acute week is inside the chronic window, the two are correlated by construction. Cutting continuous load into bands ("sweet spot", "danger zone") loses information and can create false thresholds.
  • Methods papers. Moderate. Relevance: never show the person a ratio with coloured zones; keep the estimator continuous.

3.14 Windt J, Gabbett TJ. 2017. How do training and competition workloads relate to injury? The workload—injury aetiology model. Br J Sports Med 51(5):428–435. doi:10.1136/bjsports-2016-096040 — verified (Crossref).

  • Load acts through two paths: it builds fitness (protective) and fatigue (harmful); risk depends on the person's current state, not on load alone.
  • Conceptual model. Opinion. Useful framing: the same load can be fine on a rested day and harmful after three heavy days.

3.15 Buist I, Bredeweg SW, van Mechelen W, et al. 2008. No effect of a graded training program on the number of running-related injuries in novice runners: a randomized controlled trial. Am J Sports Med 36(1):33–39. doi:10.1177/0363546507307505 — verified (PubMed 17940147).

  • RCT, n = 532 novices: a 13-week programme built on the "10% per week" rule vs a standard 8-week programme gave the same injury incidence (20.8% vs 20.3%, p = .90).
  • RCT, moderate. A fixed weekly-increase rule has no evidence behind it.

3.16 Frandsen JSB, Hulme A, Parner ET, et al. 2025. How much running is too much? Identifying high-risk running sessions in a 5200-person cohort study. Br J Sports Med 59(17):1203–1210. doi:10.1136/bjsports-2024-109380 — verified (PubMed 40623829).

  • 5205 runners, 588 071 sessions, 18 months. Injury rate rose when a single session exceeded the longest run of the past 30 days by >10% (HRR 1.64 for +10–30%, 1.52 for +30–100%, 2.28 for >100%). The ACWR showed a negative dose-response; week-to-week ratio showed nothing.
  • Large prospective cohort, observational. Moderate-to-strong for running. Relevance by analogy: the risky event is the single biggest day or task compared with the person's recent maximum, not the weekly average. A planner should cap any one day (and any one task) near the heaviest the person handled well in the last ~30 days.

3.17 Practice: TrainingPeaks Performance Management Chart (Coggan). https://www.trainingpeaks.com/learn/articles/what-is-the-performance-management-chart/ (vendor docs; consulted via secondary pages 2026-10-05).

  • Chronic load ("fitness", CTL) = exponentially weighted average of daily load with 42-day time constant; acute ("fatigue", ATL) = 7-day; "form" (TSB) = CTL − ATL of yesterday. Coaches watch the weekly rise of CTL ("ramp rate"), typically 3–7 points/week for amateurs.
  • Opinion/practice. Shows the field settled on two exponential averages (short = fatigue, long = capacity) rather than one window.

3.18 Muth JF. 1960. Optimal properties of exponentially weighted forecasts. J Am Stat Assoc 55(290):299–306. doi:10.1080/01621459.1960.10482064 — verified (Crossref).

  • Exponential smoothing is the optimal (minimum mean-squared-error) forecast when the underlying level follows a random walk and each observation adds independent noise; the best smoothing constant is set by the ratio of the two variances. This is the steady-state Kalman filter of the "local level" model.
  • Mathematical result. Strong. Relevance: if daily capacity drifts slowly and daily loads are noisy readings of it, an EWMA (or the Kalman filter it comes from) is the right family; a Gaussian kernel has no such justification and, being two-sided, needs future days unless cut in half.

3.19 Moussa I, Leroy A, Sauliere G, et al. 2019. Robust Exponential Decreasing Index (REDI): adaptive and robust method for computing cumulated workload. BMJ Open Sport Exerc Med 5(1):e000573. doi:10.1136/bmjsem-2019-000573 — verified (PubMed 31798948).

  • REDI_today = Σ αᵢ·Lᵢ / Σ αᵢ with αᵢ = e^(−λi) when day i has data and αᵢ = 0 when it is missing: an exponentially weighted mean over observed days only, renormalised. Stays consistent with large parts of the data missing, where rolling ACWR/EWMA "could not be computed with 30% missing data".
  • Simulation/methods paper. Moderate as engineering. This is the direct template for Sioul's estimator: days without tracking, and deliberate rest days, get weight 0 instead of load 0.

3.20 Benson LC, Stilling C, Owoeye OBA, Emery CA. 2021. Evaluating methods for imputing missing data from longitudinal monitoring of athlete workload. J Sports Sci Med 20(2):188–196. doi:10.52082/jssm.2021.188 — verified (PubMed 33948096).

  • 93 youth basketball players, a season of sRPE and jump counts: when a day's load is missing, the best guesses use the context of the session plus the individual; plugging in zero is not among the good methods.
  • Simulation on one cohort. Weak-to-moderate. Supports "missing ≠ zero".

3.21 Engineering analogues for a capacity that is only seen through outcomes (not health studies; cited for the mechanism).

  • Levitt H. 1971. Transformed up-down methods in psychoacoustics. J Acoust Soc Am 49(2B):467–477. doi:10.1121/1.1912375 — verified (Crossref). Thresholds are estimated from binary outcomes by stepping the stimulus up after successes and down after failures; "2-down/1-up" converges on the level passed 70.7% of the time, "3-down/1-up" on 79.4%. Transposed: raise the budget only after k consecutive days that went fine, lower it after one bad day; the budget then settles on a load the person handles on ~70–80% of days.
  • Chiu D-M, Jain R. 1989. Analysis of the increase and decrease algorithms for congestion avoidance in computer networks. Comput Netw ISDN Syst 17(1):1–14. doi:10.1016/0169-7552(89)90019-6 — verified (Crossref). Additive increase / multiplicative decrease (AIMD) is the increase/decrease rule shown to converge to efficient and fair operation, oscillating just under a capacity that is unknown and may change (it underlies TCP congestion control). Transposed: grow the budget by small fixed steps, cut it by a fraction after a crash.
  • Tobin J. 1958. Estimation of relationships for limited dependent variables. Econometrica 26(1):24–36. doi:10.2307/1907382 — verified (Crossref). The statistics of censored observations: when a value is only known to lie above (or below) what was observed, treating it as exact biases the estimate. Transposed: a day that went fine shows capacity ≥ that day's load, not = it; a day followed by a crash shows capacity < the load. Averaging all loads as if they were capacity is the biased estimate.
  • Strength: mathematical/engineering results, not evidence about people. They show the asymmetric update is a known, stable design, not an ad hoc choice.

4. Executive-function load, fatigue, effort and recovery

4.1 Ego depletion and its replication failures. Hagger MS, Chatzisarantis NLD, Alberts H, et al. 2016. A multilab preregistered replication of the ego-depletion effect. Perspect Psychol Sci 11(4):546–573. doi:10.1177/1745691616652873 — verified (PubMed 27474142). Vohs KD, Schmeichel BJ, Lohmann S, et al. 2021. A multisite preregistered paradigmatic test of the ego-depletion effect. Psychol Sci 32(10):1566–1581. doi:10.1177/0956797621989733 — verified (PubMed 34520296).

  • 23 labs, N = 2141: d = 0.04 [−0.07, 0.15]. 36 labs, N = 3531, run by the effect's proponents: d = 0.06, data 4× more likely under the null; exploratory hint of a larger effect in people reporting more fatigue.
  • Preregistered multi-site replications. Strong. The short-term "a few minutes of self-control drains a willpower tank" effect is near zero in healthy students. It does not speak to hours-long work, illness or burnout; it does mean Sioul should not model minute-by-minute "willpower spending".

4.2 Job V, Dweck CS, Walton GM. 2010. Ego depletion — is it all in your head? Implicit theories about willpower affect self-regulation. Psychol Sci 21(11):1686–1693. doi:10.1177/0956797610384745 — verified (PubMed 20876879).

  • People who believed willpower is not limited showed no depletion; manipulating the belief changed the effect; in a field study, limited-resource beliefs predicted more procrastination in demanding periods.
  • Four studies, student samples; later replications mixed (Vohs 2021 found no moderation by willpower beliefs). Weak-to-moderate. Design tension: presenting capacity as a shrinking tank could itself feed a limited-resource belief in people without an energy-limiting illness. Present the budget as a planning aid ("what fits today"), not as a fuel gauge being emptied.

4.3 Boksem MAS, Tops M. 2008. Mental fatigue: costs and benefits. Brain Res Rev 59(1):125–139. doi:10.1016/j.brainresrev.2008.07.001 — verified (PubMed 18652844). Kurzban R, Duckworth A, Kable JW, Myers J. 2013. An opportunity cost model of subjective effort and task performance. Behav Brain Sci 36(6):661–679. doi:10.1017/S0140525X12003196 — verified (PubMed 24304775).

  • Mental fatigue arises when the perceived energy cost of going on outweighs the expected reward; effort is felt as the opportunity cost of using control for this task rather than another.
  • Theory/reviews. Moderate as theory. Relevance: the same task costs less when it is wanted or rewarding, so a task's cost rating is not fixed; meaningful tasks and tasks with a visible payoff should be allowed to feel lighter, and ratings should be revisable.

4.4 Hockey GRJ. 1997. Compensatory control in the regulation of human performance under stress and high workload: a cognitive-energetical framework. Biol Psychol 45(1–3):73–93. doi:10.1016/S0301-0511(96)05223-4 — verified (PubMed 9083645).

  • Under high load people either protect performance by spending extra effort, at the price of strain and fatigue after-effects ("latent decrement"), or lower their goals. Costs can stay hidden while output looks normal.
  • Theory with experimental support. Moderate. Relevance: finishing all tasks does not prove the load was within capacity; the after-effect (evening exhaustion, the next day) is the signal. Sioul needs an outcome rating, not just task completion.

4.5 Wiehler A, Branzoli F, Adanyeguh I, Mochel F, Pessiglione M. 2022. A neuro-metabolic account of why daylong cognitive work alters the control of economic decisions. Curr Biol 32(16):3564–3575. doi:10.1016/j.cub.2022.07.010 — verified (PubMed 35961314).

  • Over roughly a workday of computer tasks, a high-demand cognitive-control group (vs a low-demand group) ended with more glutamate in lateral prefrontal cortex and shifted toward low-effort, short-delay choices.
  • One experiment (n = 40 per press reports), needs replication. Weak-to-moderate. It is the first physiological marker that cognitive fatigue accumulates over a day of demanding work and is cleared by rest/sleep, which supports a daily budget with carry-over decay rather than per-task "willpower".

4.6 Meijman TF, Mulder G. 1998. Psychological aspects of workload. In: Drenth PJD, Thierry H, de Wolff CJ (eds.) Handbook of Work and Organizational Psychology, Vol. 2: Work Psychology, pp. 5–33. Hove: Psychology Press. — exists (book chapter, no DOI; cited by hundreds of indexed papers, e.g., Geurts & Sonnentag 2006).

  • Effort-recovery model: effort at work produces short-term load reactions that are reversible if recovery follows; when recovery is insufficient, reactions accumulate and become lasting health damage. Recovery can happen during work (breaks, lighter tasks) and after work.
  • Theory. Opinion-level but the foundation of the field. Relevance: the unit to protect is the cycle load → recovery, not the load alone. A day's plan should contain the recovery for its load.

4.7 Geurts SAE, Sonnentag S. 2006. Recovery as an explanatory mechanism in the relation between acute stress reactions and chronic health impairment. Scand J Work Environ Health 32(6):482–492. doi:10.5271/sjweh.1053 — verified (PubMed 17173204).

  • External recovery (after work) is "particularly necessitated when recovery opportunities during worktime (internal recovery) are insufficient"; long hours and rumination keep activation going and block recovery; "sustained physiological activation and incomplete recovery" is the path to chronic ill health.
  • Narrative review. Moderate as synthesis. Relevance: anxiety-laden tasks are costly partly because they keep running in the head afterwards (rumination), so their cost spills into evening and night.

4.8 Sonnentag S, Fritz C. 2007. The Recovery Experience Questionnaire: development and validation of a measure for assessing recuperation and unwinding from work. J Occup Health Psychol 12(3):204–221. doi:10.1037/1076-8998.12.3.204 — verified (PubMed 17638488).

  • N = 930: four distinct recovery experiences — psychological detachment from work, relaxation, mastery (challenging, learning experiences off work), control (choosing what to do with one's time); 4 items each, built for diary studies.
  • Validation study. Strong as a measure. These four are the natural vocabulary for Sioul's "well-being gain" (section 6).

4.9 Bennett AA, Bakker AB, Field JG. 2018. Recovery from work-related effort: a meta-analysis. J Organ Behav 39(3):262–275. doi:10.1002/job.2217 — verified (Crossref). Wendsche J, Lohmann-Haislah A. 2017. A meta-analysis on antecedents and outcomes of detachment from work. Front Psychol 7:2072. doi:10.3389/fpsyg.2016.02072 — verified (PubMed 28133454).

  • 54 samples, N = 26 592: detachment after work has the strongest (negative) link with fatigue; control experiences the strongest (positive) link with vigour. 86 publications, N = 38 124: detachment goes with less exhaustion, better sleep, more well-being, small-to-medium effects; heavy work investment and after-hours work-related activity go with less detachment.
  • Meta-analyses of mostly correlational data. Strong for association. Relevance: different gains fix different deficits — detachment lowers fatigue, control raises vigour. "Gain" is not one thing either.

4.10 de Bloom J, Kompier M, Geurts S, et al. 2009. Do we recover from vacation? Meta-analysis of vacation effects on health and well-being. J Occup Health 51(1):13–25. doi:10.1539/joh.K8004 — verified (PubMed 19096200).

  • 7 studies: vacation improves health and well-being (d = +0.43) but the effect fades soon after return to work (d = −0.38). Follow-ups: de Bloom J, Geurts SAE, Taris TW, et al. 2010, Work Stress 24(2):196–216 (n = 96: back to pre-vacation levels in the first week of work); Kühnel J, Sonnentag S. 2011, J Organ Behav 32(1):125–143 (n = 131 teachers: gains in engagement and burnout gone within one month) — both from publisher/repository abstracts, DOIs not checked.
  • Small meta-analysis. Moderate. Relevance: one restful stretch does not bank capacity for weeks; recovery must be regular. It also caps how long an estimator should "remember" a holiday: the 14–28-day horizon is consistent with this fade-out.

4.11 Breaks. Albulescu P, Macsinga I, Rusu A, et al. 2022. "Give me a break!" A systematic review and meta-analysis on the efficacy of micro-breaks for increasing well-being and performance. PLoS ONE 17(8):e0272460. doi:10.1371/journal.pone.0272460 — verified (PubMed 36044424). Hunter EM, Wu C. 2016. Give me a better break: choosing workday break activities to maximize resource recovery. J Appl Psychol 101(3):302–311. doi:10.1037/apl0000045 — verified (PubMed 26375961).

  • Micro-breaks (≤10 min), 22 samples, N = 2335: vigour d = .36, fatigue d = .35, performance d = .16 (n.s.); performance gains only for low-demand tasks; longer breaks gave larger gains, so "recovering from highly depleting tasks may need more than 10-minute breaks". Experience sampling (95 employees × 5 days): breaks taken earlier in the shift and spent on preferred activities restored more; long breaks or frequent short breaks beat rare short ones.
  • Meta-analysis; ESM study. Moderate. Relevance for "heavy then light": after a cognitively heavy block, plan a real break (>10 min, of the person's choosing), early rather than late in the day; short pauses help mood, not output.

4.12 What the section implies for "a heavy task followed by a light one".

  • Supported: recovery must follow effort within the day (4.6, 4.7); cognitive fatigue accumulates over hours of demanding work (4.5); breaks restore well-being and longer breaks are needed after demanding tasks (4.11); alternating hard and easy days lowers monotony and illness in athletes (3.1); NICE asks to "alternate and vary between different types of activity" (2.6); DISC suggests recovering in the domain that was spent (1.9).
  • Not supported: minute-level willpower accounting (4.1). No trial has tested "heavy then light" task ordering as such. Ordering studies in hospitals (4.13) bear on easy-first vs hard-first, not on alternation.

4.13 Task ordering: easy-first is not neutral. KC DS, Staats BR, Kouchaki M, Gino F. 2020. Task selection and workload: a focus on completing easy tasks hurts performance. Manage Sci 66(10):4397–4416. doi:10.1287/mnsc.2019.3419 — verified (Crossref). Ibanez MR, Clark JR, Huckman RS, Staats BR. 2018. Discretionary task ordering: queue management in radiological services. Manage Sci 64(9):4389–4407. doi:10.1287/mnsc.2017.2810 — verified (Crossref).

  • Six years of emergency-department data plus a lab replication: under high workload, people pick easy tasks first, driven by fatigue and the sense of progress; after adjusting for complexity, throughput and learning are worse. 2.4 million radiology reads: picking the shortest tasks first lowers productivity.
  • Large observational datasets with lab/IV support; healthy professionals. Moderate. Relevance: left alone under load, people drift to easy tasks and hard ones pile up; a planner that places one hard task early, when fresh, counters this. It is about productivity, not health; for someone near their limit, hard-first still has to respect the budget.

4.14 Dread and time of day. Berns GS, Chappelow J, Cekic M, et al. 2006. Neurobiological substrates of dread. Science 312(5774):754–758. doi:10.1126/science.1123721 — verified (PubMed). Loewenstein G. 1987. Anticipation and the valuation of delayed consumption. Econ J 97(387):666–684. doi:10.2307/2232929 — verified (Crossref). Schmidt C, Collette F, Cajochen C, Peigneux P. 2007. A time to think: circadian rhythms in human cognition. Cogn Neuropsychol 24(7):755–789. doi:10.1080/02643290701754158 — verified (Crossref). Mark G, Iqbal ST, Czerwinski M, Johns P. 2014. Bored Mondays and focused afternoons: the rhythm of attention and online activity in the workplace. Proc CHI 2014:3025–3034. doi:10.1145/2556288.2557204 — verified (Crossref; findings not re-read).

  • Waiting for an unpleasant event has its own cost; some people accept a stronger shock to wait less (fMRI, n ≈ 32). Cognitive performance varies across the day with chronotype and time awake; attention at work has a daily and weekly rhythm.
  • Experiments; review; field study. Moderate. Relevance: anxiety-heavy tasks cost more the longer they hang over the day, so place them early in the person's good window, not at day's end; let the person declare their own good hours rather than assuming mornings.

5. Prior art in apps that plan around energy or difficulty

Searched 2026-10-05: vendor sites/help centres, app-store listings, PubMed ("Tiimo": 0 hits; planner/scheduling apps for adult ADHD with "trial": nothing evaluating a planner). No peer-reviewed outcome study of any of these planners was found. Everything below is vendor documentation (opinion level) unless stated.

5.1 Structured — "Energy Monitor". https://help.structured.app/en/articles/998530 (read 2026-10-05).

  • Each task is marked draining (three intensities), recharging, or neutral. Points ≈ per 30 min: low 1, medium 2, high 3. Default daily limit 20 points; their guide recommends 36–42 for "an average person" and 18–22 for people with limited energy (illness). Monitor turns green → yellow → red as the plan fills. Recharging tasks "only have an effect when being placed in the middle of the day, as you cannot entirely recharge your energy or more than your daily available energy limit." Built "in collaboration with researchers and medical professionals of the Dutch institute Breinstraat" (brain-injury energy management). Limit set by hand; no learning from history.
  • Closest existing product to Sioul's idea. Its choices to note: one cost dimension, intensity × duration, a hand-set cap, recharge capped and only mid-day. Sioul's additions would be three scales, a learned estimate, and a separate floor for gains.

5.2 Tiimo (visual planner, ADHD/autism). https://www.tiimoapp.com/ (read 2026-10-05).

  • AI breaks tasks into steps and estimates durations; mood/"energy and focus" check-ins to "spot patterns over time". Third-party reviews describe a "State of Mind" check-in (energy, emotional load, focus) after tasks (unverified on vendor pages). No budget, no planning against capacity; site says tools are "rooted in executive functioning research" without citations.

5.3 Sunsama (daily planning ritual). https://roadmap.sunsama.com/improvements/p/set-workload-threshold-per-individual-day (read 2026-10-05).

  • Daily "workload threshold" in hours, set by the user, can differ per weekday; warns when planned time exceeds it. Time only, no difficulty.

5.4 Reclaim.ai (calendar auto-scheduler). https://help.reclaim.ai/en/articles/12304529-auto-schedule-breaks-travel-time-with-buffers (read 2026-10-05).

  • Automatic "decompression time" after meetings (all meetings or only video calls), travel buffers, and breaks between tasks/habits; durations set by the user; nothing adapts to energy or history. The only mainstream tool that builds "light after heavy" into the calendar, keyed on event type rather than rated cost.

5.5 Motion (AI auto-scheduler). https://www.usemotion.com/help/time-management/auto-scheduling/auto-scheduling-how-to-guide (read 2026-10-05).

  • Schedules by work hours, deadlines and priority; optional fixed breaks ("a 15-minute break every 2 hours"). The help page mentions no difficulty or energy input; a third-party review claims "energy level" settings (unverified, contradicted by other reviews).

5.6 Amazing Marvin. https://amazingmarvin.com/features/all (search summary 2026-10-05, not re-read).

  • Optional task attributes include energy/focus level; lists can be sorted or filtered by energy; "strategies" limit how many tasks are shown per day to reduce overwhelm. No budget arithmetic.

5.7 Goblin.tools — Magic ToDo. Third-party descriptions (e.g., University of Warwick library guide https://warwick.ac.uk/services/library/using/productivity-tools/magic-todo-gobin-tools).

  • Breaks a task into steps; a 1–5 "spiciness" setting asks for finer breakdown when a task feels harder. No scheduling. Relevant evidence for the mechanism: unpacking a task into steps reduces the planning fallacy (7.4) and lack of structure is a source of aversiveness at the doing stage (1.15).

5.8 Llama Life. Vendor site not re-read; widely described as a timed to-do list (a countdown per task). No energy model.

5.9 Rise (sleep app with an "energy schedule"). https://www.risescience.com/blog/app-using-sleep-science-to-help-you-be-more-productive (search summary 2026-10-05).

  • Predicts hourly energy peaks and dips from circadian rhythm and "sleep debt"; sleep debt weights last night 15% and the 13 previous nights 85%, more recent heavier — a 14-night recency-weighted window. Vendor claim, method not published in a journal.

5.10 Exoplan ("energy-aware calendar"). App Store listing https://apps.apple.com/app/exoplan/id6739639089 (read 2026-10-05).

  • Predicts an energy curve from Apple Watch data after a calibration period, or from chronotype otherwise, and suggests where to place deep work vs admin. No evidence cited.

5.11 Visible and Bearable — see 2.19 and 2.20 (pacing budget from heart rate with hand-adjusted zones; 14-day biometric baseline; symptom–factor correlations after ≥3+3 days).

5.12 Published evaluations closest to this design.

  • Aboagye NY, Baker MR, Baker K, Del Din S. 2026. Wearable- and mobile app-based activity pacing and fatigue management in post-COVID-19 condition: exploratory observational study. JMIR Form Res 10:e91829. doi:10.2196/91829 — verified (PubMed 42214072). 19 adults, 2182 days, Fitbit + app: people chose pacing goals on days when already more fatigued (confounding by indication); on pacing days, "active" or "balanced" pacing went with less same-day fatigue than "light" pacing; next-day effects small and non-significant; 32% of variance between persons; 7 of 12 who tried several strategies responded. Weak (observational, small) but the only daily-data study of app-guided pacing found; it shows strong individual differences.
  • McCready JL, Campbell M, McCay K, et al. 2026. Optimising and beta-testing a user-centred, accessible, self-management rehabilitation smartphone app (reCOVer) for long-COVID fatigue. Disabil Rehabil 48(11):3425–3443. doi:10.1080/09638288.2025.2577872 — verified (PubMed 41140205). Qualitative, n = 11: an activity-pacing diary helped people see boom–bust patterns; asked for text-to-speech and clearer goals. Weak.
  • "Energy-based planning" in ADHD coaching: only blogs and coaching material found (e.g., "plan by energy cost, not time"; "front-load expensive tasks"); no peer-reviewed study located. Opinion.

6. Well-being gain: the floor to fill

The design adds, per task/event, a "well-being gain" (rest, joy, connection, movement, mastery…), with a daily minimum to reach. Sources on whether restorative activity can be scheduled, how to rate it, whether gains offset costs, and where to place it.

6.1 Behavioural activation (BA) and pleasant-activity scheduling. Lewinsohn PM, Graf M. 1973. Pleasant activities and depression. J Consult Clin Psychol 41(2):261–268. doi:10.1037/h0035142 — verified (PubMed 4147832; abstract not indexed, summary from secondary sources). MacPhillamy DJ, Lewinsohn PM. 1982. The Pleasant Events Schedule: studies on reliability, validity, and scale intercorrelation. J Consult Clin Psychol 50(3):363–380. doi:10.1037/0022-006X.50.3.363 — verified (Crossref).

  • Daily counts of pleasant activities track daily mood (n = 30: 10 depressed, 10 psychiatric controls, 10 non-depressed; within-person correlations ranged from about zero to strong, |r| up to ~.66, per secondary summaries). The Pleasant Events Schedule rates 320 activities for frequency and pleasantness (0–2 each).
  • Small observational study; psychometric study. Weak-to-moderate. Origin of "schedule pleasant activities to lift mood", and of the observation that people differ a lot in how much activities move their mood.

6.2 Dimidjian S, Hollon SD, Dobson KS, et al. 2006. Randomized trial of behavioral activation, cognitive therapy, and antidepressant medication in the acute treatment of adults with major depression. J Consult Clin Psychol 74(4):658–670. doi:10.1037/0022-006X.74.4.658 — verified (PubMed 16881773). Richards DA, Ekers D, McMillan D, et al. 2016. Cost and outcome of behavioural activation versus cognitive behavioural therapy for depression (COBRA): a randomised, controlled, non-inferiority trial. Lancet 388(10047):871–880. doi:10.1016/S0140-6736(16)31140-0 — verified (PubMed 27461440).

  • N = 241: in more severe depression BA matched antidepressants and both beat cognitive therapy. COBRA, N = 440: BA delivered by junior workers was non-inferior to CBT at 12 months (PHQ-9 8.4 vs 8.4) and cheaper.
  • RCTs. Strong. Scheduling valued/pleasant activities is an effective treatment in its own right, not a reward for finishing work.

6.3 Ekers D, Webster L, Van Straten A, et al. 2014. Behavioural activation for depression; an update of meta-analysis of effectiveness and sub group analysis. PLoS ONE 9(6):e100100. doi:10.1371/journal.pone.0100100 — verified (PubMed 24936656). Mazzucchelli TG, Kane RT, Rees CS. 2010. Behavioral activation interventions for well-being: a meta-analysis. J Posit Psychol 5(2):105–121. doi:10.1080/17439760903569154 — verified (PubMed 20539837). Also Cuijpers P, van Straten A, Warmerdam L. 2007. Clin Psychol Rev 27(3):318–326. doi:10.1016/j.cpr.2006.11.001 — verified (Crossref).

  • 26 RCTs, N = 1524: BA beats controls (SMD −0.74) and medication (−0.42, k = 4); study quality mostly low, follow-ups short. In 20 RCTs (N = 1353) BA raised well-being by g = 0.52, in non-clinical people as well as in people with depressive symptoms.
  • Meta-analyses. Strong (with the quality caveat). Evidence that a planner can legitimately schedule restorative/valued activities as a daily target.

6.4 Rating gains: established practice. CBT activity diaries rate each activity for mastery (achievement) and pleasure (enjoyment) on 0–10 (Beck AT, Rush AJ, Shaw BF, Emery G. 1979. Cognitive Therapy of Depression. New York: Guilford — book; worksheet practice confirmed at https://www.psychologytools.com/resource/mastery-and-pleasure-activity-diary). Veale D. 2008. Behavioural activation for depression. Adv Psychiatr Treat 14(1):29–36. doi:10.1192/apt.bp.107.004051 — verified (Crossref).

  • Two separate 0–10 ratings, mastery and pleasure, recorded after the activity; people in depression often report low numbers (2–3/10 pleasure), which is expected and still useful.
  • Clinical practice. Opinion-to-moderate (decades of use inside effective trials). Relevance: a single "gain" number is simpler, but the field keeps at least mastery and pleasure apart because they diverge (chores give mastery without pleasure).

6.5 Sherdell L, Waugh CE, Gotlib IH. 2012. Anticipatory pleasure predicts motivation for reward in major depression. J Abnorm Psychol 121(1):51–60. doi:10.1037/a0024945 — verified (PubMed 21842963).

  • n = 38 depressed, 30 controls: depressed people liked the rewards as much once they had them, but their low anticipatory pleasure, not their liking, predicted unwillingness to make effort.
  • Lab study. Weak-to-moderate. Relevance: a gain rated in advance will be underestimated by someone depressed or burnt out; rate gain after the activity, and plan from experienced gains (the BA principle of acting before feeling like it).

6.6 Recovery experiences as the vocabulary of gain. Sonnentag & Fritz 2007 (4.8): detachment, relaxation, mastery, control. Bennett et al. 2018 and Wendsche & Lohmann-Haislah 2017 (4.9): detachment is most tied to lower fatigue, control to vigour. Karabinski T, Haun VC, Nübold A, Wendsche J, Wegge J. 2021. Interventions for improving psychological detachment from work: a meta-analysis. J Occup Health Psychol 26(3):224–242. doi:10.1037/ocp0000280 — verified (PubMed 34096763).

  • 34 interventions, N = 3725: detachment can be trained (d = 0.36); longer and higher-dose interventions work better; effects larger in people already impaired.
  • Meta-analysis. Moderate-to-strong. Relevance: gains are learnable and dose-dependent, which supports a daily minimum rather than an occasional treat.

6.7 Sonnentag S. 2001. Work, recovery activities, and individual well-being: a diary study. J Occup Health Psychol 6(3):196–210. doi:10.1037/1076-8998.6.3.196 — verified (Crossref; summary from publisher/repository abstract). ten Brummelhuis LL, Bakker AB. 2012. Staying engaged during the week: the effect of off-job activities on next day work engagement. J Occup Health Psychol 17(4):445–455. doi:10.1037/a0029213 — verified (PubMed 22799771).

  • 100 teachers × 5 days: low-effort, social and physical activities after work raised bedtime well-being, work-related activities lowered it; high time pressure had a lagged negative effect. 74 employees × 5 days: social, low-effort and physical leisure raised next-morning vigour through detachment and relaxation; work and household duties lowered it.
  • Diary studies, healthy workers. Moderate. Relevance: an evening gain shows up as next-morning capacity, so gains belong in the estimator as a predictor of tomorrow, not as a deduction from today's costs. Household chores are costs even when "not work".

6.8 Fredrickson BL, Mancuso RA, Branigan C, Tugade MM. 2000. The undoing effect of positive emotions. Motiv Emot 24(4):237–258. doi:10.1023/A:1010796329158 — verified (PubMed 21731120).

  • After anxiety-induced cardiovascular arousal, contentment or amusement films brought heart rate and blood pressure back to baseline faster than neutral or sad films (n = 170; replication n = 185 ruled out simple replacement).
  • Lab experiments. Moderate. The best evidence that a gain can partly undo a cost within the hour: a pleasant block right after an anxiety-heavy task speeds recovery from it.

6.9 Autonomy decides whether a "break" restores. Trougakos JP, Hideg I, Cheng BH, Beal DJ. 2014. Lunch breaks unpacked: the role of autonomy as a moderator of recovery during lunch. Acad Manage J 57(2):405–421. doi:10.5465/amj.2011.1072 — verified (Crossref; summary from publisher abstract). Trougakos JP, Beal DJ, Green SG, Weiss HM. 2008. Making the break count: an episodic examination of recovery activities, emotional experiences, and positive affective displays. Acad Manage J 51(1):131–146. doi:10.5465/amj.2008.30764063 — verified (Crossref). Hunter & Wu 2016 (4.11).

  • Social and work activities at lunch increased end-of-day fatigue when imposed and reduced it when freely chosen; relaxation helped regardless. Respite activities (rest, eating, chosen socialising) during breaks raised positive affect afterwards; chores did not. Preferred activities, taken earlier in the day, restored more.
  • Experience-sampling studies. Moderate. Relevance: the planner may reserve restorative slots, but the person picks what fills them; a "social" gain proposed by the app can turn into a cost. Autistic burnout literature (2.17) makes the same point for social contact and masking.

6.10 Restorative environments. Stevenson MP, Schilhab T, Bentsen P. 2018. Attention Restoration Theory II: a systematic review to clarify attention processes affected by exposure to natural environments. J Toxicol Environ Health B 21(4):227–268. doi:10.1080/10937404.2018.1505571 — verified (PubMed 30130463).

  • 42 studies: exposure to natural environments improves working memory and cognitive flexibility (low-to-moderate effects); real exposure seems stronger than virtual.
  • Systematic review with meta-analyses. Moderate. Relevance: time outdoors is a specific gain for cognitive cost, consistent with "recover in the domain spent" (1.9).

6.11 Allostatic load. McEwen BS. 1998. Protective and damaging effects of stress mediators. N Engl J Med 338(3):171–179. doi:10.1056/NEJM199801153380307 — verified (PubMed 9428819). Juster R-P, McEwen BS, Lupien SJ. 2010. Allostatic load biomarkers of chronic stress and impact on health and cognition. Neurosci Biobehav Rev 35(1):2–16. doi:10.1016/j.neubiorev.2009.10.002 — verified (Crossref). Zijlstra FRH, Cropley M, Rydstedt LW. 2014. From recovery to regulation: an attempt to reconceptualize 'recovery from work'. Stress Health 30(3):244–252. doi:10.1002/smi.2604 — verified (Crossref).

  • Stress responses protect in the short term; wear accumulates when they are repeated, prolonged, or not switched off (lack of adaptation or recovery). Measured by multi-system biomarker indices, not by self-report. Zijlstra et al. recast recovery as regulation back toward an optimal state, not as "rest = refill".
  • Review/theory, large literature on the biomarker index. Moderate as theory; no quantitative rule for a planner. It supports two things only: recovery must be frequent (not banked), and an evening that keeps arousal high (rumination, unfinished anxiety tasks) costs even without visible work.

6.12 Do gains offset costs within a day, or need their own minimum? What the sources allow.

  • Partial offset exists for arousal: positive emotion speeds physiological recovery from anxiety within minutes (6.8); recovery activities restore well-being by bedtime and vigour next morning (6.7).
  • No source shows that gains create extra capacity for more cost on the same day. In energy-limiting illness every activity, pleasant ones included, spends energy: NICE counts "all types of activity (cognitive, physical, emotional and social)" (2.6); emotional events trigger PEM in 90% (2.8). Vacation gains fade within weeks (4.10), so gains do not bank.
  • BA treats valued/pleasant activity as a daily dose with its own target, independent of how much work was done (6.2, 6.3). Detachment works better at higher dose (6.6).
  • Conclusion (inference from the above): keep gains as a separate floor, never as a currency that buys more cost; let a met floor raise tomorrow's expected capacity slightly through the estimator (as evidence), not today's budget. Place gain right after the costliest block (6.8, 4.11) and leave evening detachment protected (4.9, 6.7).

6.13 Conflicts specific to gains.

  • The same activity can be cost and gain (social time, exercise, mastery tasks). Physical movement is a gain for most people but a PEM trigger in ME/CFS and long COVID (2.6, 2.8, 2.9) — the item must carry both a cost and a gain rating, and movement's cost must not be hidden because it is "good for you".
  • Anticipated gain is underestimated in depression (6.5) while anticipated anxiety is overestimated (1.16): both predicted ratings are biased in opposite directions, so after-the-fact ratings should drive learning.
  • Imposed restorative activity can become a cost (6.9); a gain target can become one more duty. Minimum should be small, chosen, and never scored as failure.

7. Recovery pace after burnout, time estimation, and what to show

7.1 How slowly capacity comes back. Glise K, Wiegner L, Jonsdottir IH. 2020. Long-term follow-up of residual symptoms in patients treated for stress-related exhaustion. BMC Psychol 8:26. doi:10.1186/s40359-020-0395-8 — verified (PubMed 32188513). Oosterholt BG, Maes JHR, Van der Linden D, Verbraak MJPM, Kompier MAJ. 2016. Getting better, but not well: a 1.5 year follow-up of cognitive performance and cortisol levels in clinical and non-clinical burnout. Biol Psychol 117:89–99. doi:10.1016/j.biopsycho.2016.02.009 — verified (PubMed 26930250).

  • Seven years after treatment for exhaustion disorder (n = 217; 163 clinically assessed): almost half still report fatigue, 73% reduced stress tolerance, a third still clinically exhausted. After 1.5 years, clinical burnout patients had fewer symptoms but still reported cognitive problems and showed minor test impairment.
  • Cohort follow-ups. Moderate. Recovery is measured in months to years; a planner that lets the budget climb back to "normal" in weeks is miscalibrated for this group. A slow ceiling on increases is consistent with this.

7.2 Return to work after clinical burnout. Perski O, Grossi G, Perski A, Niemi M. 2017. A systematic review and meta-analysis of tertiary interventions in clinical burnout. Scand J Psychol 58(6):551–561. doi:10.1111/sjop.12398 — verified (PubMed 29105127). Ahola K, Toppinen-Tanner S, Seppänen J. 2017. Interventions to alleviate burnout symptoms and to support return to work among employees with burnout: systematic review and meta-analysis. Burnout Res 4:1–11. doi:10.1016/j.burn.2017.02.001 — verified (Crossref).

  • 8 trials: interventions shortened time to return to work (HR 4.5, very heterogeneous) but did not change full return or exhaustion/depression/anxiety symptoms; the successful ones involved a workplace dialogue.
  • Meta-analyses of few, variable trials. Weak-to-moderate. No trial tests a particular rate of increasing workload after burnout; the rate is a clinical judgement.

7.3 Kratz AL, Murphy SL, Braley TJ. 2017. Ecological momentary assessment of pain, fatigue, depressive, and cognitive symptoms reveals significant daily variability in multiple sclerosis. Arch Phys Med Rehabil 98(11):2142–2150. doi:10.1016/j.apmr.2017.07.002 — verified (PubMed 28729168).

  • 107 adults, 0–10 ratings 5×/day for 7 days: fatigue and pain vary a lot within and between days; fatigue rises across the day.
  • EMA study. Moderate. Relevance: day-to-day noise is large, so a single day says little; estimates need ≥1–2 weeks of days, and daily plans need margins.

7.4 Planning fallacy and time estimation. Buehler R, Griffin D, Ross M. 1994. Exploring the "planning fallacy": why people underestimate their task completion times. J Pers Soc Psychol 67(3):366–381. doi:10.1037/0022-3514.67.3.366 — verified (Crossref). Roy MM, Christenfeld NJS, McKenzie CRM. 2005. Underestimating the duration of future events: memory incorrectly used or memory bias? Psychol Bull 131(5):738–756. doi:10.1037/0033-2909.131.5.738 — verified (PubMed 16187856). Kruger J, Evans M. 2004. If you don't want to be late, enumerate: unpacking reduces the planning fallacy. J Exp Soc Psychol 40(5):586–598. doi:10.1016/j.jesp.2003.11.001 — verified (Crossref).

  • People underestimate how long their tasks will take; one reason is that they also underestimate how long past tasks took; listing the sub-steps of a task reduces the underestimate.
  • Experiments and review. Strong for the bias, moderate for the unpacking fix. Relevance: plan from tracked durations, not stated ones; leave slack; breaking tasks into steps also improves the estimate.

7.5 ADHD and time. Zheng Q, Wang X, Chiu KY, Shum KK. 2022. Time perception deficits in children and adolescents with ADHD: a meta-analysis. J Atten Disord 26(2):267–281. doi:10.1177/1087054720978557 — verified (PubMed 33302769). Mette C. 2023. Time perception in adult ADHD: findings from a decade — a review. Int J Environ Res Public Health 20(4):3098. doi:10.3390/ijerph20043098 — verified (PubMed 36833791).

  • 27 studies (1620 with ADHD, 1249 controls): less accurate (g > 0.40) and less precise (g = 0.66) timing in young people with ADHD. In adults, studies are few and mixed.
  • Meta-analysis (youth); narrative review (adults). Moderate / weak. Relevance: tracked time is more trustworthy than estimates for ADHD users; margins should be larger when estimates are used.

7.6 What not to show: evidence on measurement, framing and streaks.

  • Etkin J. 2016. The hidden cost of personal quantification. J Consum Res 42(6):967–984. doi:10.1093/jcr/ucv095 — verified (Crossref). Six experiments: measuring an enjoyable activity increased how much people did but reduced how much they enjoyed it, "can make enjoyable activities feel more like work", and lowered continued engagement and well-being. Moderate. Direct risk for gain ratings: rating restorative activities can make them less restorative.
  • Silverman J, Barasch A. 2022. On or off track: how (broken) streaks affect consumer decisions. J Consum Res 49(6):1095–1117. doi:10.1093/jcr/ucac029 — verified (Crossref). Seven studies: highlighting a broken streak lowers subsequent engagement compared with an intact one, more so when people blame themselves; being able to "repair" the streak softens it. Moderate. No streaks in Sioul.
  • Ancker JS, Witteman HO, Hafeez B, et al. 2015. "You get reminded you're a sick person": personal data tracking and patients with multiple chronic conditions. J Med Internet Res 17(8):e202. doi:10.2196/jmir.4209 — verified (PubMed 26290186). 22 patients, 7 providers: tracking "feels like work"; health numbers "provoke strong positive and negative emotions, value judgments". Qualitative, weak-to-moderate. Keep rating burden low and numbers out of sight by default.
  • Wohl MJA, Pychyl TA, Bennett SH. 2010. I forgive myself, now I can study: how self-forgiveness for procrastinating can reduce future procrastination. Pers Individ Dif 48(7):803–808. doi:10.1016/j.paid.2010.01.029 — verified (Crossref; summary from secondary reports). 119 students: forgiving oneself for procrastinating before the first exam predicted less negative affect and less procrastination before the second. Weak-to-moderate. After a bad or unproductive day, neutral or kind wording helps the next day; blame does not.
  • Counter-evidence: Kim YH, Jeon JH, Choe EK, Lee B, Kim K, Seo J. 2016. TimeAware: leveraging framing effects to enhance personal productivity. Proc CHI 2016:272–283. doi:10.1145/2858036.2858428 — verified (Crossref). 24 participants, 8 weeks: only those shown their distracting time (negative framing) improved productivity. Weak (small, healthy knowledge workers, outcome = productivity, not health). Noted as a conflict: negative framing can push output up in healthy users; Sioul's aim and users differ, and 7.6's other sources point the other way.

Design criteria

Plain rules for Sioul, each with its sources. "Inference" marks a rule that combines sources rather than following one directly; numbers marked "default" are starting values to tune, not findings.

Scales and ratings

  1. Keep cognitive and emotional cost as separate scales: occupational health measures them separately with little overlap, and matched recovery works best within the same domain — 1.9 de Jonge & Dormann 2006, 1.10 COPSOQ 2005, 1.3 Galy 2018.
  2. Keep anxiety/aversiveness as its own scale, because it predicts postponement independently of difficulty and its cost runs before (dread) and after (rumination) the task — 1.14 Steel 2007, 1.15 Blunt & Pychyl 2000, 4.14 Berns 2006, 4.7 Geurts & Sonnentag 2006.
  3. Anchor the emotional scale on having to hide or fake feelings, not only on sad or tense content — 1.13 Hülsheger & Schewe 2011, 2.17 Raymaker 2020.
  4. Offer an optional fourth "body and senses" scale (physical, sensory, standing), or let people fold it into the closest scale, since these loads trigger crashes in energy-limiting illness and in autistic burnout — 2.6 NICE 2021, 2.8 Chu 2018, 2.17 Raymaker 2020, 2.21 #MEAction guide, 1.9 DISC (physical).
  5. Rate intensity "while doing it" on 0–10 with words at 0, 5 and 10, and let the app multiply by duration itself, because people's overall ratings ignore duration — 1.8 Fredrickson & Kahneman 1993, 3.2 Foster 2001, 1.17 Hjermstad 2011.
  6. Ask for a quick rating right after a task when possible, use it for learning, and use the planned rating only until an after-task rating exists — 1.6 van Gog 2012, 1.7 Schmeck 2015, 1.16 Rachman 1994.
  7. Learn each person's planned-versus-felt gap per scale and correct planned ratings with it (anxiety is usually over-predicted; gains under-predicted in low mood) — 1.16 Rachman 1994, LaFreniere & Newman 2020, 6.5 Sherdell 2012.
  8. Treat a difference of 1 point as noise and ignore changes under about 2 points or 30% — 1.17 Farrar 2001.
  9. To rank tasks heavy or light, use the highest of the three per-hour ratings, each scaled by that scale's budget (a task is as heavy as its worst side) — inference from 1.7 Schmeck 2015, 1.8 Redelmeier & Kahneman 1996, 1.11 Wickens 2008.

A day's load

  1. Load per scale = Σ (after-task rating × tracked hours) over tasks and calendar events of the day, the session-RPE construction — 3.2 Foster 2001, 3.3 Haddad 2017.
  2. Charge every item a minimum duration (default 30 min) so short anxiety-heavy items such as a phone call are not counted as almost free — inference from 4.14 Berns 2006, 4.7 Geurts & Sonnentag 2006; Structured counts in 30-min units (5.1).
  3. Mark a day as "missing" when too little of it was tracked (default: under half the planned items have a status), and as "rest" when the person planned rest; neither counts as low load — 3.19 Moussa 2019, 3.20 Benson 2021.

The estimator (budget per scale)

  1. Estimate capacity, not habit: a moving average of all past loads describes what the person did, so learn only from days whose outcome is known — 3.12 Impellizzeri 2021, 3.11 Impellizzeri 2020, 4.4 Hockey 1997, 2.16 Kluger 2013.
  2. Collect one outcome per day, cheap to give (evening: fine / too much; next morning: rested / not), because finishing tasks does not show the load was safe — 4.4 Hockey 1997, 2.3 O'Connor 2019, 2.19 Visible.

  3. Give missing and rest days weight zero and renormalise the remaining weights, never count them as zero load — 3.19 Moussa 2019 (REDI), 3.20 Benson 2021. 15a. Treat a light day that went fine as uninformative (capacity is at least that load, not equal to it): only good days that used most of the budget (default ≥70% on that scale) feed the average, so light weeks cannot ratchet the budget down — 3.21 Tobin 1958 (censoring), 3.12 Impellizzeri 2021, inference. 15b. Combine as follows: the budget tracks the weighted mean of those informative good days, but rises no faster than criterion 19 allows and falls only through criteria 20–21 — inference from 3.19, 3.21, 2.6.

  4. Use a one-sided exponential kernel (EWMA/REDI) rather than a Gaussian: it is the optimal estimator for a slowly drifting level seen through noisy days, needs no future days, and is the field standard — 3.18 Muth 1960, 3.9 Williams 2017, 3.10 Murray 2017, 3.7 Kolossa 2017.

  5. Default time scale: a 14-day effective window (λ ≈ 0.13/day, half-life ≈ 5 days) over at most the last 28 days; 7 days is too easily dragged by a multi-day crash — 2.19 Visible (moved 7 → 14 days), 5.9 Rise (14 nights), 3.9 Williams 2017 (7/28), 4.10 de Bloom 2009 (gains fade in weeks).
  6. Do not fit per-person decay constants; keep the structure fixed and learn only the budget level per scale — 3.5 Hellard 2006.
  7. Raise a budget only after a stable stretch (default: ≥7 included days, no bad day, average use ≥80% of budget), by a small step (default ≤10% per week), never on a timetable — 2.6 NICE 1.11.2/1.11.13–14, 2.9 WHO protocol in Parker 2023 (≥7 days per phase), 3.15 Buist 2008 (fixed 10% rule no benefit), 3.21 Levitt 1971 / Chiu & Jain 1989 (additive increase).
  8. Lower a budget at once by a fraction (default −20%) when a bad outcome follows a day at or above about 80% of budget, attributing it to the heaviest of the previous 1–3 days — 2.6 NICE (PEM delayed hours to days), 2.8 Chu 2018, 3.21 Chiu & Jain 1989 (multiplicative decrease).
  9. After a flare (default: ≥3 bad days in 7), freeze increases and re-establish the baseline once stable, accepting that it may stay lower — 2.6 NICE 1.11.15–16 and glossary ("relapses can lead to a long-term reduction").
  10. Cap any planned day, and any single task's intensity, at about the heaviest well-tolerated level of the last 30 days plus 10% — 3.16 Frandsen 2025 (single-session spikes), 3.6 Busso 2003.
  11. Cap one day's influence on the estimate (winsorise a day's load at 1.1–1.2 × current budget) so one big good day cannot lift the budget — inference from 19 and 22.
  12. Before 7 included days, plan from a starting budget the person sets, defaulting low (about half of a recent "boom" day), and give the learned estimate full weight only from about 14 days — 2.21 #MEAction guide, 2.6 NICE 1.11.4, 7.3 Kratz 2017 (large daily variability), 2.20 Bearable (≥3+3 days, 30 recommended).
  13. Expect recovery from burnout to take months to years and keep the slow increase rule even when days go well — 7.1 Glise 2020, Oosterholt 2016.
  14. Treat gains as a predictor of tomorrow's state, not as an addition to today's budget — 6.7 Sonnentag 2001, ten Brummelhuis & Bakker 2012, 6.12.

The planner

  1. Never plan above the estimate on any scale, and keep a margin (default: plan to 85% of each budget) for estimation error, daily variability and unplanned demands — 7.4 Buehler 1994, Roy 2005, 7.3 Kratz 2017, 7.5 Zheng 2022, 4.4 Hockey 1997.
  2. Keep a fourth, total budget learned the same way from the sum of the three scales, and check the day's sum against it, because the pools overlap and a day may not fill all three at once — 1.11 Wickens 2008, 1.12 Richards & Gross 2000, Eysenck 2007 (inference).
  3. Follow every heavy block with a light task or a break longer than 10 minutes, and never put two heavy blocks back to back — 4.6 Meijman & Mulder 1998, 4.11 Albulescu 2022, 2.6 NICE 1.11.4 ("plan periods of rest and activity"), 3.1 Foster 1998.
  4. Alternate the kind of cost (after a cognitively heavy block, something cognitively light even if emotionally neutral; after an emotional one, something restoring in that domain) — 1.9 de Jonge & Dormann 2006, 2.6 NICE ("alternate and vary").
  5. Put the day's hardest and most anxiety-laden items early in the person's own good hours, which the person declares — 4.13 KC 2020, 4.5 Wiehler 2022, 7.3 Kratz 2017 (fatigue rises across the day), 4.14 Berns 2006, Schmidt 2007.
  6. Break heavy or vague tasks into steps before scheduling them, which lowers aversiveness and corrects time estimates — 7.4 Kruger & Evans 2004, 1.15 Blunt & Pychyl 2000, 5.7 Goblin.tools.
  7. Plan from tracked durations, not stated ones, with wider slack for people who report time-estimation trouble — 7.4 Roy 2005, 7.5 Zheng 2022, Mette 2023.
  8. Lighten the day before and the day after an unavoidable heavy calendar event — 2.21 #MEAction guide ("radical rest"), 2.6 NICE 1.11.4 (pre-emptive rest), 2.8 Chu 2018 (delayed PEM).
  9. For energy-limiting illness keep daily totals steady (no "hard days"); for others, varying day loads across the week is acceptable — 2.6 NICE, 2.13 Andrews 2012 (boom–bust), 3.1 Foster 1998 (monotony).
  10. When the person says today is bad, re-plan today to a lighter level they choose (default ~60% of budget, or rest), move the non-urgent rest, and record the day as an outcome without lowering the long-term budget unless the previous days were heavy — 2.6 NICE 1.11.2 ("experts in judging their own limits"), 1.11.15, 2.13 Murphy & Kratz 2014 (pacing only once symptoms are already high did not help, so lighten the rest of the day at once), criterion 20.
  11. An optional morning check-in ("lighter / usual") may lower today's plan but never raise it above the estimate — 2.3 O'Connor 2019 (available-energy rating), 2.19 Visible (morning score).

Gains (the floor to fill)

  1. Give each item an optional gain rating 0–10, recorded after the activity, with anchors drawn from mastery, pleasure, connection, detachment and calm — 6.4 Beck 1979 / Veale 2008, 4.8 Sonnentag & Fritz 2007, 6.5 Sherdell 2012.
  2. Let one item carry both costs and a gain (social time, exercise, absorbing work), and never hide the cost of movement because it is "good for you" — 6.13, 2.6 NICE, 2.8 Chu 2018.
  3. Keep gains as a separate daily floor, not a currency: gains do not buy extra cost today — 6.12, 2.6 NICE (all activity costs energy), 4.10 de Bloom 2009 (no banking), 6.2–6.3 BA as a daily dose.
  4. Set the floor low at first (default: one or two restorative blocks a day) and learn it from the gain level on days the person rated good — 6.3 Mazzucchelli 2010, 6.6 Karabinski 2021 (dose matters), inference.
  5. Place a restorative block right after the costliest block of the day, and protect an evening stretch for detachment — 6.8 Fredrickson 2000, 4.9 Bennett 2018, 6.7 ten Brummelhuis & Bakker 2012, 4.11 Hunter & Wu 2016 (earlier breaks restore more).
  6. The app may reserve restorative slots, but the person picks what fills them — 6.9 Trougakos 2014, 2.17 Raymaker 2020.
  7. Keep gain rating sparse and optional (category defaults, occasional prompts), because measuring enjoyable activities makes them feel like work — 7.6 Etkin 2016.

What to show, and never show

  1. Show the plan and whether it fits ("room left today"), with the reason for any change in plain words ("yesterday was heavy, so today is lighter") — 2.6 NICE (person-led), 7.6 Ancker 2015.
  2. Never show capacity as a score, a percentage of "normal", a comparison with pre-illness or with others, or a red "over budget" alarm after the fact — 7.6 Ancker 2015, Wohl 2010, 3.13 Carey 2018 (no coloured zones).
  3. No streaks, badges, or counts of failed or postponed tasks — 7.6 Silverman & Barasch 2022, Wohl 2010.
  4. Show changes in the estimate only when larger than noise (~30%), and let the person overrule the estimate at any time — 1.17 Farrar 2001, 2.6 NICE ("experts in judging their own limits").
  5. Present the budget as "what fits today", not as a fuel tank being drained, for people whose limits are not illness-imposed — 4.2 Job 2010 (limited-resource beliefs), 4.1 Hagger 2016.
  6. Offer, but do not push, the "felt lighter than expected" comparison for anxiety-rated tasks, as exposure-based therapy does — 1.16 LaFreniere & Newman 2020, Rachman 1994.
  7. Make no health claim: the evidence for pacing apps is thin, and the app should present itself as a planning aid — 2.11 Abonie 2020, 2.7 PACE/Wilshire 2018, 5.12 Aboagye 2026.

Open questions / conflicts

  1. Habit or capacity. The proposed weighted average of past loads measures habit; sports science found the "chronic load" adds nothing beyond noise (3.12). Outcome ratings fix this but add a daily question. Unknown: the cheapest outcome signal that still works (one evening tap? next-morning tap? both?), and the share of the budget above which a good day counts as informative (70% in criterion 15a is a guess).
  2. Kernel and window. Exponential vs half-Gaussian matters less than the time scale. Evidence: 7 days too twitchy (Visible), 14 days used by two vendors, 28 days for sports chronic load, months for burnout recovery. Per-condition defaults are untested.
  3. How far back to blame. PEM is delayed 0–72 h (11% consistently ≥24 h; 2.8); burnout and ADHD costs show up the same day or overnight. A 1–3-day attribution window is a guess; it should be a setting.
  4. Ratings before or after. After-task ratings are more accurate (1.6, 1.16) but cost attention; Etkin (7.6) warns that measuring itself changes experience. How often to prompt is unresolved.
  5. Intensity × duration vs a cost per item. sRPE multiplies by duration (3.2); dread and rumination make short anxiety items expensive (4.14, 4.7); peak-end memory ignores duration (1.8). The 30-min minimum charge is a guess.
  6. Three scales or four. DISC, COPSOQ, NICE and the PEM literature all count physical/sensory load (1.9, 1.10, 2.6, 2.8); the brief has none. The anxiety scale may also be partly a procrastination-aversiveness scale rather than anxiety proper (1.14, 1.15).
  7. Combining scales. No study validates a cognitive + emotional + anxiety composite. Per-scale budgets plus a total check (criteria 9, 28) are inference; TLX research is split on global vs per-dimension scores (1.2, 1.3).
  8. Hard first or easy first. Hard-first protects throughput and learning (4.13) and reduces dread (4.14); people under load drift to easy tasks partly for the sense of progress (4.13), and a small easy start can help task initiation in ADHD (coaching practice, no trial found). Compromise in criterion 32 (a small first step of the hard task) is untested.
  9. Hard and easy days vs steady days. Sports supports varied days (3.1); pacing for energy-limiting illness wants steady days and no booms (2.6, 2.13). Criterion 35 splits by condition; who decides which applies is open.
  10. How fast to rise. No trial gives a safe rate for cognitive/emotional load after burnout. NICE forbids fixed increments; WHO stays ≥7 days per phase; a fixed 10% weekly rule did nothing in runners (3.15) while single-day spikes did matter (3.16). The ≤10%/week step is a guess.
  11. Fuel-tank framing. In healthy people, believing willpower is limited may create depletion (4.2); in ME/CFS limits are real (2.6). The same interface serves both; framing may need to differ.
  12. Shame vs nudges. Negative framing raised productivity in healthy workers (7.6 TimeAware); the target users and the goal differ, and other sources (Wohl, Silverman, Ancker) point to harm. Chosen side: no shaming displays; not tested in this population.
  13. Gains: offset or floor. No source shows gains increase same-day capacity; some show partial undoing of arousal (6.8) and next-day benefits (6.7). The floor model (criterion 40) is inference.
  14. Movement. A gain for most, a crash trigger in ME/CFS and long COVID (6.13). Default rating of exercise items must not assume gain.
  15. Weak evidence base for pacing itself. Pacing reduces fatigue modestly (2.11), PACE found its version of pacing ineffective (2.7), energy envelope studies are small and cross-sectional (2.2–2.4). Sioul builds on consensus (NICE) and engineering, not on proven efficacy; it could log anonymous, local, per-person outcome data so the person can judge for themselves.
  16. Individual differences. Between-person variance is large (5.12: ICC 0.32; 6.1: mood–activity link from none to strong). Population defaults will be wrong for many; per-person learning is limited by noisy, short histories (3.5). Only the budget level per scale should be learned at first.
  17. Weekday structure. If work days and free days differ, one budget per scale may be too coarse; excluding rest days removes most of the weekly cycle, but people with different weekday loads might need weekday-specific budgets. Untested.
  18. Margin size. 85% is a default without a source; planning-fallacy studies show the direction, not the size of a safe margin for this use.