Skip to content

Free-Choice Learning Environment Design

Principle · All principles
Evidence · 16 claims (6 for, 9 mixed, 1 against) · 18 studies (5 review, 4 theoretical, 3 causal, 3 qualitative, 2 quant-synthesis, 1 design), q1–q4 · 2 of 18 report an effect size · 14 claims rest on one study

Conditional relationship

This page owns the open-environment case: settings where the learner decides whether to come, what to do and how long to stay, with no imposed curriculum, sequence or grade (a museum floor, a science centre, a hobby community, an ungraded weekly studio block inside a school). For a learner in such a setting, an environment that offers several entry points at different levels of complexity, starts from what the learner already knows and cares about, attaches no grades or expected tangible rewards to taking part, and responds to what the learner does is expected to produce longer and more frequent voluntary engagement (persistence, motivation-affect), and, for some learners, self-extended projects that go beyond what the activity was designed to teach. Learning of a particular target (conceptual-understanding, near-transfer) is a separate outcome: it is expected only where the activity also gives feedback on the target idea, not from free exploration alone.

The relationship is conditional on the learner's starting interest and task knowledge (a learner with neither may sample and leave), on whether anything in the environment responds to an attempt (a person, a peer, the exhibit itself), on how many decisions the learner is left to make, and on which outcome counts. No claim in the wiki tests a free-choice environment as a whole against a structured or compulsory one. The nearest evidence is one randomised comparison inside a museum exhibit (responsive feedback against the same materials without it), two meta-analyses on rewards and later free-choice engagement, one school-fixed-effects study of field trips, and qualitative cases.

How the converted siblings sit beside this: Autonomy holds bounded choice inside a fixed course and the case against controlling incentives; this page applies that model to environments with no fixed course at all and does not repeat it. Engagement and Motivation hold what raises participation in required work. Self-Directed Learning holds when learner-directed study helps and for whom, and the Self-Directed Learning pattern its sequence; this page is about the environment an institution builds, not the learner's planning. Game-Based Learning covers a game whose winning move is the target reasoning, which is one way an exhibit can respond to the visitor.

Default design, while the relationship is untested

The earlier page's guidance and that of the in-school STEAM infrastructure page folded into it, kept as a concrete default a designer can act on and revise. Each step is a design proposal; its evidence status is stated beside it.

  1. Give every exhibit or challenge a first action that needs no reading and no prior knowledge, finishable in two to five minutes, and at least two deeper levels behind it. The first loop is what decides whether a passer-by stays. Untested proposal; the earlier page asked for "multiple entry points and complexity levels". In the one comparison in the wiki, average visits to an unfacilitated hands-on exhibit lasted about 1.5 minutes and to the responsive version about 6.25 minutes (AI-mediated feedback in hands-on exhibits [~M]), which is the range a first loop has to work in.
  2. Open from the learner's own interests and knowledge. Pose the starting question in everyday terms (what will make this tower fall, what do you want your game to do), and let the learner choose the topic, pace and whether to work alone or with others. Treat an interest as a starting point to go deeper from, not a fixed topic to cater to (IDC's interest cycle [+W], a theoretical commentary). Otherwise untested here.
  3. Attach no grade and no expected tangible reward to taking part; give information about what happened instead. Badges, points or prizes for merely doing an activity the learner already finds interesting are the case where later free-choice engagement fell in a 128-experiment meta-analysis, while a competing 96-experiment meta-analysis found the effect minimal and confined to expected tangible rewards for doing a task; both found verbal praise did not undermine it (expected tangible rewards and free-choice engagement [~M]). A literature-based argument from maker education reaches the same caution about grades (grading in maker education [+W]).
  4. Make the activity respond to each attempt, then fade the response. Where the outcome of a learner's action can be sensed or shown (a tower falls or stands, a circuit lights, code runs), ask for a prediction, show the result, and offer a short choice of explanations on contrasting cases that each isolate one idea; after a few cases, switch to open building where the activity only reports success or failure. Evidence: with grade-2 children in a museum, this cycle beat the same physical materials with static signage on explaining the principle and on a later unguided building task, with no loss of enjoyment (AI-mediated feedback in hands-on exhibits [~M]). One study, one topic (structural stability); carrying it to other content is untested.
  5. Design for people explaining to each other. Make the exhibit or challenge one that a group works at together, with physical materials and a shared display, so visitors talk about why something worked. A report on one museum game-exhibit says learning typically happened when visitors explained to each other, and that physical blocks plus a screen representation were needed for deeper practices (Oztoc exhibit [+W]). For group tasks, prefer a well-defined goal and materials everyone can handle (well-defined tasks and accessible materials in a museum workshop [~M], one group of four adolescents).
  6. Inside a school, protect a recurring ungraded block and stock it. The source case ran 90 minutes a week for a school year, with about 30 leveled challenges across STEAM domains, each with optional tutorials, and shared tools (an online asset repository, a 3D printer) learners could use without asking. These are the source case's values, not tested doses. Two students in one such studio extended a single challenge into a year-long project, spending over half their scheduled time on it and solving 23 technical problems the challenge never taught (choice-rich infrastructure [~M]); the authors chose them as an extreme case, and most students did not go this far.
  7. Facilitators observe and drop in; they do not direct. Train adults to circle the room, answer direct questions, and send a learner who asks for help to a peer who has gone further on that challenge before stepping in themselves. Do not check for "on-task" compliance. Untested proposal, from the folded STEAM infrastructure page; without the role being redefined, the block tends to slide back into teacher-directed instruction (the earlier page).
  8. Treat extending, combining or changing a challenge as a sign the environment is working. Allow learners to leave the designed task, and make the tools for going further reachable. Same single case as step 6 (choice-rich infrastructure [~M]).
  9. End each visit or session with a one-minute look back: what did you find out, what would you try next. Learners left to their own problem-solving often skip checking what a failed attempt taught them, and often abandon projects (youth skip the assess phase [~W]; minimal persistence in everyday problem-solving [-W], one qualitative study of urban high-school youth). The prompt is an untested proposal; keep it spoken or optional, not a form to complete.
  10. Plan an arc across sessions where learners come back. Proposal, untested: in the first two sessions, a five-minute tour of the challenge library and tools and a level-1 start for everyone, with the facilitator checking in on each learner once; in middle sessions, learners choose their own challenges and the facilitator checks in only when asked or when someone has been stuck for about ten minutes; from mid-year, learners who have finished a challenge's top level act as the peer expert for it and may start their own extension. A learner moves up a level when they have finished the current level's product without a facilitator's help. In a museum visited once, the arc is inside one visit: guided cases first, then open building (step 4).
  11. Judge the environment by what learners chose to do, not by coverage. Record dwell time, return visits, which levels were reached, which projects went beyond the designed task, and what learners could explain afterwards; there is no curriculum to measure completion against (the earlier page). Untested as a method here.

When a condition is not met: if nothing in the environment can respond to an attempt and no adult is free, use exhibits whose own result shows success or failure and add a printed contrast ("try a wide base and a narrow one"); if learners arrive with no interest in the topic, a free-choice setting may not reach them at all, and the design question becomes the opening hook rather than the depth behind it.

Fitting the design to a situation

The three facts that most change the decision:

  • Is attendance voluntary, and how often do learners come back? A one-off museum visit, a weekly school block that learners must attend but may spend as they choose, and a hobby community joined for years need different first loops and different arcs. If unstated, ask: will this learner be here once, weekly, or whenever they like, and can they leave at any moment?
  • What, if anything, responds to an attempt? A responsive exhibit, a facilitator, peers who know more, or nothing at all. The one controlled comparison in the wiki is between a responsive exhibit and the same materials without a response. If unstated, ask: when a learner tries something, who or what shows them whether it worked and why?
  • Which outcome the institution needs. Dwell time and return visits, a self-extended project, an explained idea, and a standardised test score are different outcomes, and the evidence for each is different. If unstated, ask: what would show, a month later, that the environment did its job?
If the brief says… Then change… Basis
A museum or science-centre exhibit, visited once, families or mixed ages Make the first loop a two-to-five-minute action with no reading; add a responsive predict–show–explain cycle where the outcome can be sensed; put the depth in optional later levels AI-mediated feedback in hands-on exhibits [~M] (grade-2 children, one museum exhibit on stability; mixed-age families untested)
A recurring in-school block learners must attend but may spend as they choose Protect it from grading; stock a leveled challenge library and shared tools; train facilitators to observe and route help to peers; follow the arc across sessions above Choice-rich infrastructure [~M] (one studio, two sixth graders, an extreme case; shows what is possible, not how often)
Learners are novice in the topic, or have low confidence or low literacy Keep the first action wordless; narrow the choice to two or three challenges at first; offer a worked starter project to copy and change; let a facilitator or peer start alongside Moderate agency in game-based learning [~W] (second-hand report of one game study: moderate agency beat none and high); connectivist courses [~W] (open online courses, second-hand: not all could direct themselves). Both carried to this setting
Learners are advanced or long-standing enthusiasts Point them straight to the open or highest level and to the tools for building their own; invite them to be the peer expert for a challenge untested proposal
Learners are child (primary age) Use the responsive cycle on contrasting cases before open building; keep explanations as a short menu to choose from, not free writing AI-mediated feedback in hands-on exhibits [~M] (median grade 2)
adult learners in work settings or hobby communities Start from a problem the learner brings from work or their hobby; offer resources and peers to ask rather than a sequence; agree with the learner what a finished project looks like untested proposal; no claim here tests adults in free-choice settings
The goal is understanding a particular idea, not only engagement Add feedback on that idea inside the activity (prediction, result, chosen explanation, contrasting cases), then fade to open building; check with one unaided task at the end AI-mediated feedback in hands-on exhibits [~M] (explanation posttest and unguided building task, same day)
A school is sending classes on field trips to raise science achievement Treat a visit as a small addition, not a replacement for science teaching; expect the largest gains for students with least access otherwise; record which students went and how often Field trips and middle-school science scores [+M] (New York City, six years of school data, small effects); effects largest for Hispanic and lunch-eligible students [~M] (same study)
The institution requires grades, credits or points Keep any grade off taking part; if credit is needed, give it as a flat amount for participation rather than per piece of output, and give feedback on what the learner made Expected tangible rewards and free-choice engagement [~M] (laboratory experiments; the flat-rate moderator comes from one meta-analysis and is unchecked); grading in maker education [+W]
The school is under heavy test-preparation pressure Protect the block explicitly in the timetable and say what it is for; do not justify it by test-score gains it has not been shown to produce Low-income schools and test prep [~W] (a theoretical commentary on US schools)
Groups of adolescents work at one table or exhibit Give a well-defined goal and materials everyone can handle; watch whether ideas are taken up on merit or overridden socially, and step in when one voice is shut out Well-defined tasks and accessible materials [~M] (one group of four, a museum architecture workshop)
One facilitator for many learners, or scarce facilitator time (volunteers, floor staff on a busy shift) Make the exhibits answer the learner themselves (step 4); route help to peer experts first; limit the adult to brief check-ins with those stuck for about ten minutes untested proposal; from the folded STEAM page
Mixed modes: some learners join remotely or on phones only Put the challenge library and tutorials online and phone-readable; let remote learners share their products to the same repository and ask peer experts in a shared chat; keep the first action possible without special hardware untested proposal
Single session only One exhibit or challenge, a guided first loop then open building, and a one-minute look back; record dwell time and what learners could explain; do not claim lasting interest untested proposal
A fixed short time budget (for example a 45-minute visit slot) Five minutes' welcome and choice, about 30 minutes at one or two chosen exhibits, five minutes' look back, five minutes' slack; do not try to cover the floor untested proposal
Learners' products go to a real audience Provide a public place to share (a repository, a showcase evening) and let learners choose whether to share; record downloads or visitors as a signal, not a grade Choice-rich infrastructure [~M] (the two students' games were uploaded publicly and downloaded; one case)
Private or sensitive material (a personal project, family history, identity work) The learner decides who sees it; sharing is opt-in; facilitators do not display work without asking untested proposal
The brief names its own categories, routine, rubric or criteria (its own check-in, challenge levels, evaluation form) Keep them; change only what the evidence says matters here: no grade on taking part, a response to each attempt, a look back at the end untested proposal (wave lessons)
The brief wants a whole project or performance kept whole (a game, a build, a show) Do not split it into worksheets or items; check progress by what the learner can show and explain at the project untested proposal

Ten of the twenty rows cite a claim. Three of them (the museum exhibit, child and target-idea rows) rest on one randomised comparison at one exhibit, and the field-trip row on one school-data study with fixed effects; the others rest on qualitative cases, second-hand reports or arguments carried to settings they did not study, as each row says. The other ten rows are untested proposals.

Observation, state and explanation

What can be observed is behaviour under stated conditions: whether a learner stops at an exhibit or opens a challenge, how long they stay, what they try, what they say to others, whether they come back, which level they reach, and whether they change the task. "Is interested", "is curious" and "is learning" are inferred states. Dwell time is behaviour, not understanding, and a long stay can be confusion as well as engagement.

Response Plausible explanations Observation that could change the interpretation
Leaves an exhibit within a minute The first action was unclear; the topic does not connect to anything they care about; the group they came with moved on Watch where the hand goes first and whether they read; ask one visitor what they thought it was for; compare dwell for visitors alone and in groups.
Stays a long time but cannot say what happened or why Enjoying the action without noticing the idea; the exhibit gives no feedback on the idea; the explanation needs words they do not have Ask for a prediction before the next attempt; offer a choice of explanations rather than asking for one; compare with a visitor who used the responsive version.
Repeats the same challenge level week after week Comfort and enjoyment; the next level looks too hard; no one has shown them what is behind it Show the next level with a peer who has done it; ask what they would like to make; note whether they move up when a starter is offered.
Abandons a self-started project halfway A knowledge gap they could not get past; someone else solved it for them; competing activities Ask what stopped them; check whether a peer expert or tutorial for that problem exists; see whether they return when the gap is filled.
Changes the challenge into something else entirely A self-extended project (the environment working); avoidance of a hard step; copying a friend Ask what they are trying to make and what they need next; check whether the new work uses the skills of the original challenge plus new ones.

These rows are proposals, untested as diagnostics. Asking a visitor a question changes the visit, so record each probe.

Evidence and qualifications

No claim in the wiki compares a free-choice environment with a structured or compulsory one on engagement and learning together. The claims below test parts of the relationship.

Learners, settings and outcomes differ across these claims (grade-2 children at one exhibit, laboratory reward experiments, two sixth graders in a studio, city-wide school data, high-school youth at home), so do not rank them by effect label. None measures delayed learning from a free-choice environment.

The evidence by question

Each question below links a general claim that gathers the studies behind it into one table, one row per distinct study, with each study's design and effect as recorded and no averaged verdict. The count is of distinct studies across the claim and the narrower claims that bear on it.

Further evidence, not yet read against this model

Claims bearing on the page that the model above does not use as core evidence. Each says what it is and how far it bears.

Objective and learner-valued goal

The institution's objective may be visits and dwell time, interest in a field, understanding of a particular idea, or (for schools) achievement. The learner's goal is often something else: to have fun with family, to make a thing that works, to show friends, to get better at a hobby. In a free-choice setting the learner's goal decides whether they stay, so elicit it early ("what would you like to make or find out?") and record it beside the designer's objective. Agreement and divergence are both observations: a child building the tallest tower may value height while the exhibit's objective is base width, and a contrasting case that serves both (a tall tower that stands because its base is wide) is better than overriding either.

For example, the two sixth graders in the FUSE case valued making a game other people would play; the designed challenge's objective was level design. Their goal carried them to skills the challenge never taught, and public downloads were the signal they cared about. Recording only challenge completion would have missed what they learned.

What would revise this model?

The expectation should weaken if comparisons of free-choice environments with structured ones, on comparable learners, show no difference in return visits or self-extended work, or show that responsive feedback lengthens visits without improving what learners can explain at a delay. If rewards for taking part turn out not to reduce later voluntary engagement in museums or studios (rather than laboratories), step 3 becomes a preference. If wide-open choice works as well for novices as narrowed choice, the low-confidence row should go. If responsive exhibits of the kind in the one controlled study fail on content other than structural stability, step 4 is a topic-specific design. Do not protect the model by calling every short visit a design failure or every long one engagement.

Dwell time, voluntary return, self-extended projects, explained understanding at the visit, delayed understanding and test scores are separate claims. The present evidence sets no dose for a studio block, no number of challenges and no threshold for moving up a level.

Claims

  • Learner Choice — a related but distinct concept: bounded choice offered inside an otherwise fixed course, versus free-choice design's assumption of no externally imposed curriculum at all
  • Informal Learning — free-choice learning is a specific, environment-design instantiation of the broader informal-learning concept
  • Self-Directed Learning: when learner-directed study helps and for whom; the pattern holds its sequence
  • Engagement and Motivation: participation in required work

Examples

  • The Exploratorium's Active Prolonged Engagement exhibits, which let visitors test their own theories and lead their own experiments
  • A museum exhibit offering multiple entry points and complexity levels so visitors with different prior knowledge can all engage meaningfully
  • FUSE Studio: ~30 leveled STEAM challenges, no completion requirements, a peer-expert help-seeking structure, and facilitators who "circle the room, observe, and drop in."
  • Intelligent Mixed-Reality Exhibit: the design behind the one randomised exhibit comparison (step 4)
  • Design assessment as information, not reward and punishment, and provide an engaging curriculum with choice and community
  • Co-Census — museum software that turns visitors' answers about identity, family and work into visualisations they control by moving, to spark dialogue among visitors

Key Sources

  • Ashton, S., Nelson, K. R., & Millward, L. (2018). Careers in museum learning: Making a difference in museums through instructional design. In R. West (Ed.), Foundations of Learning and Instructional Design Technology. EdTech Books. https://edtechbooks.org/lidtfoundations/careers_in_museum_learning
  • Falk, J. H., & Dierking, L. D. (2013). The museum experience revisited. Routledge.
  • Hilppö, J., & Stevens, R. (2024). Learning Inside the School, but Outside the Curriculum: An Extreme Case of Interest-Driven Learning in Alternative STEAM Learning Infrastructure for Schools. Cognition and Instruction, 42(4), 482-504. https://doi.org/10.1080/07370008.2024.2386956
  • Yannier, N., Crowley, K., Do, Y., Hudson, S. E., & Koedinger, K. R. (2022). Intelligent science exhibits: Transforming hands-on exhibits into mixed-reality learning experiences. Journal of the Learning Sciences, 31(3), 335-368. https://doi.org/10.1080/10508406.2022.2032071 (via its claim page)
Page metadata
Field Value
type principle
id free-choice-learning-environment-design
title Free-Choice Learning Environment Design
description In settings where learners choose whether, what and how long to engage, several entry points, a start from their own interests, no grades or expected rewards for taking part, and an activity that responds to each attempt are expected to lengthen voluntary engagement and, with feedback on the target idea, to improve understanding; no claim tests a free-choice environment as a whole, and one randomised exhibit study tests the feedback part.
status review
moved_from pattern/alternative-in-school-steam-learning-infrastructure
canonical True
generated by claude/unspecified on 2026-10-15
trust unverified — no one has confirmed this page's content
sources 1 — hilppo-stevens-2024

Frontmatter is the page's machine-readable record — see the Schema & Guide.