The 85 % Rule, Not Gamification
Wilson et al. (2019, Nature Communications) established the optimal-difficulty band that drives learning. Gamification mechanics are a different lever — and one that often actively harms students with disabilities.
By Davit Janunts, M.Ed. Special Education (Lehigh University — Fulbright Foreign Student Program); co-author, Morin, Janunts, et al. (2024), Exceptional Children, 90(2), 145-163, doi:10.1177/00144029231165506.
Summary
Wilson, Shenhav, Straccia & Cohen (2019, Nature Communications, 10, 4646) derived the “85 % rule” from the mathematics of stochastic-gradient-descent learning: the optimal training difficulty for a learning agent is the one at which the agent is correct approximately 85 % of the time. The result holds across machine-learning models and humans, and it explains independent classroom and laboratory findings on optimal challenge level. The implication for adaptive practice in special education is direct: the algorithmic problem to solve is “keep this student in the 85 % band on their current skill,” not “motivate this student with points and streaks.” Performance Factor Analysis (Pavlik, Cen & Koedinger 2009) and half-life-regression spaced retrieval (Settles & Meeder 2016) are the field-validated algorithms for staying in that band. Gamification — points, streaks, badges, leaderboards — is a separate mechanism that recruits a separate motivational system, and the peer-reviewed evidence is that for many students with disabilities, that system competes with learning rather than supporting it.
The cost shape — what gamification actually adds in a SPED context
Sweller (1988, Cognitive Science, 12(2), 257-285) cognitive-load theory predicts that any on-screen element competing for working memory with the target skill subtracts from learning. Streak counters, point totals, badge progress bars, and leaderboard positions all qualify. For a typically-developing student with abundant working-memory capacity, the cost is small. For a student with a math-learning disability, executive-function impairment, or processing-speed deficit, the same element consumes a larger share of available working memory at the moment of the difficult task. Ashcraft & Krause (2007, Psychonomic Bulletin & Review, 14(2), 243-248) document that math anxiety operates through the same working-memory channel — the anxious student loses capacity on the task, and any additional surface element accelerates the loss.
The 85 % rule itself does not require any of these elements. The algorithm needs only to choose the next item, observe the response, and update its estimate of skill. Nothing in Wilson et al. (2019) requires that the student see a point counter.
Three categories where 85 % rule and gamification diverge
1. Motivational displacement — extrinsic rewards undermine intrinsic motivation
Deci, Koestner & Ryan (1999, Psychological Bulletin, 125(6), 627-668) is the authoritative meta-analysis of 128 experimental studies on the effect of extrinsic rewards (points, badges, tokens, monetary prizes) on intrinsic motivation. The headline finding is durable: tangible expected rewards reliably reduce intrinsic motivation for interesting tasks, especially when the reward is contingent on engagement or completion. The 85 % rule produces a different reinforcer — the intrinsic satisfaction of working at the correct difficulty level — and it does so without recruiting the extrinsic-reward circuit. The difficulty-management problem is solved by the algorithm, not by the points display.
2. Feedback distortion — evaluative praise vs. task/process feedback
Hattie & Timperley (2007, Review of Educational Research, 77(1), 81-112) is the most-cited synthesis of feedback effects in education. The headline result: feedback at the task and process levels produces effect sizes of d=0.95-1.10 on learning. Feedback at the self level — “Great job!”, “You’re a genius!”, badges and stars — produces d=0.14 (small). Most gamification mechanics deliver self-level feedback by construction: the badge celebrates the student, not the work. Replacing “Correct! +10 XP” with “You used the part-whole strategy on this fraction problem” shifts feedback from self-level to process-level and recovers the order-of-magnitude effect difference.
3. Anxiety amplification — visible failure is the modal experience for SPED students
Streak loss, leaderboard rank-drops, and visible “0 / 5 correct” counters externalize the difficulty experience the SPED student already lives with. Ashcraft & Krause (2007) document the working-memory cost of math anxiety; the same working-memory channel is implicated in reading anxiety and test anxiety. Gamification elements that surface failure publicly recruit the anxiety response and reduce the available capacity for the task that is supposed to be the unit of learning. The 85 % rule, by contrast, is a private algorithmic state — the student experiences only that the items are at the level they can handle.
Why the 85 % rule is sufficient — and what to add instead of gamification
The 85 % band is not the only piece of the adaptive-practice algorithm. Pavlik, Cen & Koedinger (2009) Performance Factor Analysis estimates the student’s skill on each tracked target and chooses the next item to maintain the band. Settles & Meeder (2016) half-life regression schedules retrieval practice across sessions to maximize long-term retention; Carpenter, Pan & Butler (2022, Nature Reviews Psychology, 1, 496-511) review the broader spaced-retrieval evidence base. Kalyuga (2007, Educational Psychology Review, 19(4), 509-539) expertise-reversal-effect research adds the constraint that scaffolds should fade as skill grows; forcing scaffolds on high-skill students harms learning. None of these algorithmic ingredients requires gamification; all of them produce the underlying learning gains gamification is sometimes credited with.
Macnamara & Burgoyne (2023, Psychological Bulletin, 149(5-6), 329-354) report a pre-registered meta-analysis (N=97,672) finding that the academic effect of growth-mindset interventions is essentially zero (d=0.02) in the highest-quality studies. The cautionary lesson is the same one the 85 %-vs-gamification distinction makes at the algorithmic level: psychological levers that look like they should work often do not, while the boring algorithmic ingredients (right difficulty, right spacing, faded scaffolds, task-level feedback) reliably do.
Equity guard — gamification harms compound at the margin
The students for whom gamification is most often justified — students who are “not motivated,” who “need extra incentive” — are also the students for whom the cognitive-load, feedback-distortion, and anxiety-amplification costs documented above are largest. Working-memory deficits, processing-speed differences, and prior experiences of academic failure are not evenly distributed; they cluster within IDEA-eligible categories and are well-documented in the SPED research literature. A general-education gamification design that produces a small net benefit for typically-developing students can produce a net cost for the SPED-eligible subgroup the same product is supposed to serve. The cleaner equity policy is the one this brief argues for: solve the difficulty-management problem with the algorithm; let the points layer go.
What changes operationally
The shift is from “what makes the student come back tomorrow” (gamification framing) to “what makes the time the student spends today produce learning” (algorithmic framing). Operationally that is a series of yes/no questions on the product surface: Is the current item targeting an 80-90 % success band for this student on this skill? Is the next-item decision driven by a PFA-equivalent estimator on the right skill? Does the feedback after a response describe what the student did, or evaluate the student? Is the scaffold tier appropriate for the student’s current PFA, or is the system pushing scaffolds at a high-skill student? Are streaks, badges, leaderboards, or visible point totals on screen during the task itself? Removing the last category, and ensuring the first four are answered correctly, is the work.
The deliverable is not a more motivating screen. It is a correctly-tuned screen that needs less motivation by construction.
Disclaimer. This brief is a research-informed analysis of published peer-reviewed cognitive-science, motivation-psychology, and education-research literature. It is not a clinical recommendation for any specific student or product. Districts and product teams evaluating gamification mechanics should consult their instructional-design lead and the relevant peer-reviewed primary sources cited inline.
References
- Ashcraft, M.H., & Krause, J.A. (2007). Working memory, math performance, and math anxiety. Psychonomic Bulletin & Review, 14(2), 243-248.
- Carpenter, S.K., Pan, S.C., & Butler, A.C. (2022). The science of effective learning with spacing and retrieval practice. Nature Reviews Psychology, 1, 496-511.
- Deci, E.L., Koestner, R., & Ryan, R.M. (1999). A meta-analytic review of experiments examining the effects of extrinsic rewards on intrinsic motivation. Psychological Bulletin, 125(6), 627-668.
- Hattie, J., & Timperley, H. (2007). The power of feedback. Review of Educational Research, 77(1), 81-112.
- Kalyuga, S. (2007). Expertise reversal effect and its implications for learner-tailored instruction. Educational Psychology Review, 19(4), 509-539.
- Macnamara, B.N., & Burgoyne, A.P. (2023). Do growth mindset interventions impact students’ academic achievement? A systematic review and meta-analysis with recommendations for best practices. Psychological Bulletin, 149(5-6), 329-354.
- Pavlik, P.I., Cen, H., & Koedinger, K.R. (2009). Performance Factors Analysis — A new alternative to knowledge tracing. Proceedings of the 14th International Conference on Artificial Intelligence in Education (AIED 2009), 531-538.
- Settles, B., & Meeder, B. (2016). A trainable spaced repetition model for language learning. Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics (ACL), 1848-1858.
- Sweller, J. (1988). Cognitive load during problem solving: Effects on learning. Cognitive Science, 12(2), 257-285.
- Wilson, R.C., Shenhav, A., Straccia, M., & Cohen, J.D. (2019). The Eighty Five Percent Rule for optimal learning. Nature Communications, 10, 4646.