A learning app notices that a student has missed three sessions. It could explain the value of practice, offer a smaller task and let the student change reminders. It could also announce that everyone else is keeping up, conceal the unsubscribe control and send a late-night warning. Both designs may produce another login. Only one makes it easy for the student to decide why, whether and how to return.
Motivation without manipulation is the work of helping people act on reasons they can understand and endorse. This does not mean that every action must be enjoyable or that influence can be eliminated. Teachers set assignments, clinicians make recommendations, managers allocate tasks and interfaces arrange choices. The ethical and empirical question is how those arrangements affect a person’s capacity to understand options, participate in a decision and revise it when circumstances change.
Motivation is more than an amount
A person can work hard because the task matters to them, because they want a reward, because they fear punishment or because they would feel ashamed to refuse. Counting minutes worked would make these motives look similar. Their experience and likely durability may differ. Self-determination theory describes a distinction between more autonomous forms of regulation and more controlled forms, as well as the needs for autonomy, competence and relatedness that can shape engagement.
Autonomy in this literature is not the same as doing everything alone. A student may willingly follow a demanding teacher’s plan because the reason is clear and the student endorses the goal. A worker may accept a safety rule without finding the task intrinsically interesting. Conversely, an interface can offer twenty superficially different buttons while making refusal confusing or costly. The number of options is an unreliable measure of agency.
Deci and Ryan’s account of autonomy support and control argues that social contexts can support chosen, self-endorsed action or pressure people toward prescribed outcomes. It is a theoretical framework informed by research, not a test that labels every reminder or reward as harmful. The practical implication is to look at the whole relationship: what reasons are offered, what feedback says, how disagreement is handled and whether the person can genuinely change course.
Support three needs without pretending they are buttons
Autonomy support starts with a meaningful rationale and an honest account of constraints. A teacher might say why an exercise matters, acknowledge that it can be frustrating and invite students to choose a topic or pace where that choice is real. A supervisor might explain a non-negotiable deadline while allowing workers to shape the route to it. This is different from saying “you choose” when the only acceptable answer has already been selected.
Competence support means a person can understand the task, receive usable feedback and make progress at a workable level of difficulty. A streak counter may record repeated action, but it does not explain what was learned. Feedback that names a specific next step can be more informative than praise that merely demands enthusiasm. Relatedness support concerns whether people feel respected and connected, not whether they are exposed to social comparison or public rankings.
A 2024 systematic review and meta-analysis by Wang and colleagues examined self-determination-theory-based interventions in education. It reviewed 36 intervention studies involving 11,792 participants and reported evidence of benefits for motivational and need-related outcomes, with variation across intervention features and measures. The review supports taking autonomy-supportive design seriously; it does not supply a universal script for every class, workplace or app. The details of implementation and the setting remain important.
The three needs can also be in tension with institutional demands. An organization may tell staff to pursue “meaningful” work while leaving unsafe workloads untouched. An app may celebrate mastery but reward only continuous use. Support is credible when it changes conditions as well as messages. A motivational phrase cannot compensate for missing time, accessibility, fair treatment or a feasible task.
Influence is unavoidable; manipulation is a design risk
Every presentation of a choice has an order, a default, a deadline and some amount of friction. Even a blank form asks its questions in a sequence. The presence of a choice architecture therefore does not by itself prove manipulation. A prompt can make an important option easier to notice, or a default can reduce effort for a choice many people would have made anyway. The question is whose interests the arrangement serves and how readily a person can understand and resist it.
Mertens and colleagues’ meta-analysis of choice-architecture interventions found behavioral effects across the studied domains, but with substantial heterogeneity. Its evidence addresses whether interventions shifted measured choices, not whether every shift was freely endorsed. A successful click-through rate cannot answer whether users understood the consequence, regretted the decision later or could reverse it without a penalty.
The paper also has a published correction addressing a retracted study and data or coding errors, and its publication-bias interpretation drew formal debate. That history is a reason to use the synthesis as evidence that interventions have been tested across domains, not as a promise of one reliable effect size. A local intervention still needs a comparison that measures its own outcomes and possible costs.
Manipulation becomes a stronger concern when a designer hides material information, makes one path disproportionately hard, exploits moments of vulnerability or presents a false social norm. A payment page that makes a paid subscription prominent but buries cancellation uses more than ordinary explanation. A health reminder that claims certainty where evidence is limited asks people to act on a distorted picture. The exact ethical judgment depends on context, stakes and evidence of how people experience the design.
In a registered experiment, Turetski and colleagues varied the type of choice intervention, its domain and the rationale used to describe it in hypothetical scenarios. Participants’ ratings of acceptability and perceived autonomy varied by intervention type and domain. The explanation’s effect was not a simple overall main effect in the full study, although some interactions suggested that framing could matter in particular combinations. These were judgments about described interventions, not observations of long-term real-world behavior. Their value here is to challenge the idea that “nudges” are either always acceptable or always coercive.
What a respectful prompt can do
Consider a reminder for a voluntary course. It can say what the next session covers, how long it takes and why the learner might find it useful. It can offer a shorter alternative, a pause and a clear way to turn reminders off. These choices do not guarantee that the learner returns. They make the decision legible. If the learner declines, the system can respect that decision rather than escalating pressure automatically.
For a clinician, a respectful prompt may explain the expected benefit and uncertainty of a recommended action, invite questions and make a follow-up plan possible. It must fit clinical duties and the patient’s circumstances. For a manager, it may connect a task to its purpose, allow questions and acknowledge where choice is constrained by a genuine requirement. An ethical account should not disguise employment power as a casual preference.
Choice should be proportionate to the decision. A learner need not configure every detail of a lesson to have agency. Too many poorly explained options can burden a person instead of helping them. A clear default accompanied by visible alternatives and an easy reversal may be more supportive than a maze of settings. What matters is whether the available choice corresponds to a real difference the person can understand and use.
Feedback can focus on information rather than leverage. “You completed three examples; the fourth still needs a different method” gives the learner something to work with. “Only weak students stop now” uses identity threat. A team might measure whether the first form of feedback improves understanding and sustained participation, rather than treating any increase in session length as proof of better motivation.
There is also a difference between an invitation and a demand disguised as an invitation. A message saying “Your learning plan is waiting if you want to resume” describes an available opportunity. A message saying “Your future self will be disappointed unless you return tonight” tries to recruit guilt. Either message may attract attention; only an evaluation that asks about understanding, pressure and later participation can reveal more than the immediate opening rate. Even then, responses can vary between people. A reminder that feels welcome to one learner may feel intrusive to someone balancing care work or illness.
Designers can give users control over timing and channel rather than assuming a more forceful message is the next step when engagement falls. The system might offer a weekly digest, a pause until a chosen date or no prompts at all. It can record those preferences without treating silence as a failure. For a course, the institution should ask whether its schedule and materials are accessible before assigning the entire problem of missed sessions to individual motivation.
When the person says no
The strongest test of a supposedly voluntary design is often what happens after refusal. Can someone pause notifications without a guilt screen? Can a patient ask for a second explanation? Can a worker raise a concern without losing access to essential information? A nominal opt-out with a hidden path, repeated confirmation or punitive language may preserve a formal choice while making its exercise expensive.
Consent is not a single click when a system continues to learn from behavior and change its prompts. A person may agree to a weekly reminder but not a sequence of escalating messages inferred from their late-night use. A meaningful design lets them see what will happen, change the frequency and end the intervention. It also treats vulnerable situations with more care, since stress, dependence and power differences can affect how free a choice feels.
There are limits to personal choice. A school may require attendance, a workplace may require safety practice and public health services may have obligations that cannot be reduced to individual preference. The honest version of autonomy support names those constraints and gives people a voice over what remains open. Inventing a choice where there is none is itself misleading.
Measure quality of motivation, not just compliance
A program can track behavior, but it should also ask what the behavior means. Useful measures might include whether people understand the rationale, feel able to disagree, can perform the task independently and continue when prompts stop. In a learning setting, an assessment of understanding or transfer may matter more than daily opens. In a workplace, retention, strain and perceived fairness may reveal costs hidden by a short-term productivity number.
Comparison matters. A new reminder system may coincide with a campaign, a new instructor or a change in available time. An evaluation can compare different versions of a prompt and include a condition with no extra prompt. It should report who was included, who ignored or disabled the intervention and whether the benefit persists. If the intervention works mainly for people already motivated, claiming it solves a broader engagement problem would mislead.
People can also report feeling pressured while producing the desired behavior. That response is not an inconvenience to be averaged away. It is part of the outcome. The design may improve a narrow metric while reducing trust or widening a gap for people with less time, different language needs or lower digital confidence. A respectful program should be willing to revise its target when the metric conflicts with the person’s welfare or stated purpose.
Interpretation needs a time horizon. A prompt might lift participation for one week and then lose its effect as people habituate. A participant may also learn to wait for prompts instead of planning independently. Neither outcome is inevitable, but both are plausible questions for follow-up. If the aim is durable learning, assess whether people can use the skill after the reminder ends. If the aim is an informed one-time decision, test whether they can explain the option and its consequences later.
For high-stakes decisions, the test should extend beyond one screen. Someone may select a default quickly, discover an unexpected cost a month later and struggle to undo it. Follow-up can examine comprehension, regret, reversal, continued engagement and any unequal burden. None of these alone settles the ethical question, but together they give a fuller account than conversion.
What the evidence does not show
The cited education review does not show that every autonomy-supportive practice works in every age group or setting. The 1987 autonomy account is a framework, not proof that all rewards undermine intrinsic motivation. Some external structures are useful and can be endorsed by the person who follows them. The relevant question is how they are introduced and experienced.
The nudge meta-analysis does not establish that a changed choice is an informed or durable choice. Nor does the registered acceptability experiment establish that a positively rated hypothetical intervention will have acceptable effects when deployed at scale. Ratings, behavior, understanding and welfare are different outcomes. A design should not use evidence about one as a substitute for evidence about the others.
There is no validated universal checklist that turns all influence into ethical motivation. The suggestions here combine research findings with an Institute synthesis about practice: state the real purpose, disclose material constraints, make alternatives usable, keep refusal and reversal possible, and measure more than compliance. These principles require judgment in the particular domain and should be revised when people affected by the design report a problem.
Return to the learning app
The student’s fourth login is useful information, but it is not the whole story. Did the app help them choose a manageable exercise, or did it make them anxious about losing a streak? Did they learn something they wanted to learn, and can they stop the reminders tomorrow? Those questions connect design performance to the learner’s purpose.
Motivation without manipulation is not a promise to remove all influence. It is a discipline of making influence intelligible, answerable and proportionate. A person should be able to see what is being asked, understand why, receive support to act competently and still remain the author of a meaningful yes or no.
Research references
- Deci, E. L., & Ryan, R. M. (1987). The support of autonomy and the control of behavior. Journal of Personality and Social Psychology, 53(6), 1024–1037. View source
- Wang, Y., Wang, H., Wang, S., et al. (2024). A systematic review and meta-analysis of self-determination-theory-based interventions in the education context. Learning and Motivation, 87, 102015. View source
- Mertens, S., Herberz, M., Hahnel, U. J. J., & Brosch, T. (2022). The effectiveness of nudging: A meta-analysis of choice architecture interventions across behavioral domains. Proceedings of the National Academy of Sciences, 119(1), e2107346118. View source
- Turetski, D., Rondina, R., Hutchings, J., Feng, B., & Soman, D. (2023). An experimental investigation into whether choice architecture interventions are considered ethical. Scientific Reports, 13, 18334. View source
- Deci, E. L., Olafsen, A. H., & Ryan, R. M. (2017). Self-determination theory in work organizations: The state of a science. Annual Review of Organizational Psychology and Organizational Behavior, 4, 19–43. View source
- Correction for Mertens et al. (2022). The effectiveness of nudging: A meta-analysis of choice architecture interventions across behavioral domains. Proceedings of the National Academy of Sciences, 119(19), e2204059119. View correction
- Maier, M., Bartoš, F., Stanley, T. D., Shanks, D. R., Harris, A. J. L., & Wagenmakers, E.-J. (2022). No evidence for nudging after adjusting for publication bias. Proceedings of the National Academy of Sciences, 119(31), e2200300119. View critique