DRAFT v0.1 Philosophy · Political epistemology

Freedom to Learn

Millian experimentation on a dancing landscape

A free society needs room to try different lives. It also needs the capacity to learn from them while the world changes around it.

This position, stated plainly

A companion to True As Of.

Claim
The freedom to experiment needs institutions that preserve the capacity to learn.
Scope
How liberal institutions support discovery and justify the uses made of experience.
Standing
Individual agency has value independently of the knowledge it produces for others.
Status
An argument and a research program, open to criticism and revision.
As of

How to read this page. The main column develops the argument. The margins draw out consequences for institutional design. The five conditions are followed by objections, limits, and the evidence the position still needs.

The freedom to try is the beginning

A society can contain a great many experiments and learn remarkably little from them. People try different arrangements, announce their successes, copy one another, and move on. There is activity, sometimes considerable prosperity, and no shortage of confident explanations. None of this establishes how much has been learned, which lessons will travel, or whether the people affected have had a meaningful say.

I take Mill’s defense of experiments in living seriously. Choosing a life develops the person who chooses it; encountering different lives enlarges the possibilities others can consider. Mill also understood that inherited experience can be misinterpreted or poorly suited to another person’s circumstances. His argument gives individuality value beyond its contribution to a social research program. [1]

My question concerns an additional defense of experimentation: that a diverse, experimental society will discover better ways of living together. Under what conditions can that claim be sustained? The modern world makes the question pressing because information can circulate quickly, practices can spread widely, and the consequences of one experiment can change the circumstances of another.

My position is that liberty needs an institutional environment in which experience can remain informative. The relevant freedom includes room to attempt something, time to understand it, and standing to revise or reject it. A successful experimental order preserves those capacities even when its participants cannot see the whole system.

Muldoon already makes the world move

Ryan Muldoon develops Millian experimentation into an account of political discovery. Experiments can disclose possibilities and considerations of justice that an established framework does not yet recognize. His 2015 paper explicitly addresses a changing landscape; his later book treats social arrangements as subject to repeated bargaining and revision. Justice becomes a trajectory. It would be a mistake to present social change as an objection his account had forgotten. [2] [3]

My contribution begins with a further feasibility question. A theory that relies on discovery must account for the capacities discovery requires. In my thesis, I argue that an account of legitimacy through public justification assigns citizens a tracking task: they need ways to recognize when changing circumstances alter what institutions do and whether their authority remains justified. That burden can be redistributed through institutions, but it cannot simply be assumed away. [4]

The same scrutiny applies to experimental liberalism. More alternatives do not automatically produce reliable observations, responsible interpretations, or acceptable terms of cooperation. My argument is intended as an extension of the experimental project: specify its enabling conditions, examine where they fail, and make their maintenance part of institutional design.

What moves on a dancing landscape?

A useful practice occupies a particular social setting. Its consequences depend on the people involved, their expectations, available resources, surrounding institutions, and what other people are doing. Those relationships can change. More significantly, the practice itself can change them. Success attracts imitation; imitation alters incentives; altered incentives change the meaning of success.

My workshop paper calls attention to this combination of endogenous change and limited perception. Individuals and institutions work with partial representations and delayed feedback. They can respond sincerely to a world that has already moved. The paper proposes local experimentation and distributed feedback as ways to extend individual capacities, without claiming that any observer can hold a complete map. [5]

Three kinds of movement need to be distinguished. We can acquire better evidence about what happened. The circumstances supporting a result can change. Experience can also change what participants consider valuable. Discovering an overlooked cost, losing a favorable background condition, and reconsidering one’s purposes are different events. Each calls for a different explanation.

In True As Of, I distinguish truth from warranted assertion and ask that a claim specify its domain, tolerance, test record, and date. The political application adds a question about transfer: what supports using this result here, for these people, now? A past success can remain a fact even when a present recommendation needs to be revised. [6]

One experiment, several different claims

Consider a hypothetical group of firms trying a four-day working week. After a year, they report stable output and improved retention. Suppose the observations are accurate. What have they established?

They have evidence about those firms, their employees, and the arrangements they actually adopted. Generalizing requires more work. Participation may have selected unusually capable managers or jobs amenable to scheduling changes. Other firms may face different customer demands. Some benefits might depend on complementary practices that a short description leaves out.

There is another complication. Part of the recruitment advantage may arise because the arrangement is unusual. If every employer adopts it, that particular advantage disappears. Other benefits may survive, and new ones may emerge as transport, childcare, and customer expectations adjust. Wider adoption changes the question being tested.

None of this discredits the original experiment. It identifies the different claims we might make from it: the arrangement benefited these participants; something similar may benefit others; widespread adoption may improve the institutional environment. Each claim needs support appropriate to its scope.

Some practices succeed as components of a diverse system. Their contribution may depend on coexistence with alternatives. Experimental learning can therefore improve the terms on which different lives fit together without showing that everybody should live the same one.

There are at least three clocks

Consequences take time to emerge. Interpretation takes time. Institutional revision takes time. Those delays must be compared with the period during which the relevant circumstances remain sufficiently similar. This comparison is a diagnostic idea, not an established numerical law.

When circumstances change slowly enough for informative outcomes to emerge and be understood, experience can accumulate into useful guidance. When they change sooner, a society can adopt yesterday’s conclusion after its applicability has weakened. More observations do not solve the problem if their interpretation depends on relationships that no longer hold.

Communication speed is only one part of the learning process. An account of a school reform can circulate in an afternoon. Its effects on durable learning may take much longer to establish. A new organization can report early enthusiasm before its effects on training, trust, or succession become visible. The announcement can be fast while the knowledge remains unfinished.

The appropriate response is to match the pace of revision to the phenomenon. Some experiments need early intervention because their costs are substantial. Others need protection from repeated interruption. An institution that changes direction with every new signal can make its own experience impossible to interpret. Responsiveness includes knowing when sufficient evidence has not yet arrived.

Connection helps. So does independence.

Modern communication can help unusual people find one another, share experience, and sustain alternatives that would struggle in isolation. It can also make nominally separate experiments depend on the same examples, assumptions, and measures. A thousand implementations may repeat one source of error.

The institutional issue is how to preserve enough independence for discovery and enough exchange for learning. Neither maximum connection nor maximum separation supplies a general answer. In work with Keith Hankins and Alexander Schaefer, Muldoon models how communication and search procedures condition the benefits of diversity. Their results depend on the problem and the way participants work together; a procedure useful in one setting can be counterproductive in another. [7]

Laboratory research offers a useful illustration, with appropriate limits. Bernstein, Shore, and Lazer found that intermittent interaction in a problem-solving task combined benefits of exchange with continued exploration. This is evidence for studying the timing of interaction, not a demonstration that democracies should operate like laboratory teams. [8]

Public visibility introduces a further pressure. An experiment can become a political symbol before its consequences mature. Participants may acquire reputational reasons to defend a success story or conceal a disappointment. Institutions that support learning need space for candid accounts, internal criticism, and revision without treating every changed judgment as a confession of bad faith.

The same questions apply when automated systems help people generate alternatives or interpret results. Lowering the cost of inquiry could broaden participation. Shared defaults could also narrow the questions being asked. Several fluent summaries of the same evidence are not several independent tests. The relevant assessment concerns what the tools enable people to notice, compare, and challenge, including the assumptions carried by the tools themselves.

The conditions of experimental learning

I propose the following principle: an institution appealing to experimentation as a liberal justification owes two accounts: how experience can improve judgment, and how those involved or affected retain their standing as agents.

Five conditions make that commitment concrete. Some concern the reliability of learning; others concern the terms on which it occurs. Accurate knowledge can be obtained through unjust treatment, and free choices can yield little useful knowledge. Each requirement needs its own defense. These are revisable design commitments whose implementation will differ across domains.

Agency

Participants have meaningful choices, opportunities to criticize, and practical routes to leave or revise an arrangement. Effects on people outside the experiment remain subject to justification.

Informative feedback

There are credible ways to notice consequences, including delayed costs and inconvenient results. The evaluation can expose failure rather than merely certify compliance with its own assumptions.

Independent alternatives

People can sustain different approaches and question prevailing measures of success. Exchange informs experiments without requiring immediate convergence on the most visible example.

Responsible transfer

Claims about wider adoption identify relevant differences in context, scale, participants, and effects. Local success does not carry an unlimited license to generalize.

Continuing revision

People can challenge conclusions and the procedures used to reach them. Institutions preserve records, detect consequential changes, and remain capable of correction.

The principle concerns a claim made on behalf of an institutional process. A person choosing an unfamiliar life owes society no experimental protocol. Where authorities invoke social learning to justify rules or expose others to consequences, however, the promised learning should be open to examination.

Whose success, by which standard?

There is a limitation in the landscape metaphor that I want to make explicit. People can have incomplete views of a common problem. They can also understand the same consequences and value them differently. Those forms of disagreement should not be collapsed into one.

One person may prefer more independence at the cost of some security. Another may reasonably choose the reverse. Experience can clarify the consequences without settling the ranking. Gaus raises this issue in his exchange with Muldoon: people need not share one evaluative landscape, so a discovery for one community need not constitute an improvement everyone can endorse. [9]

A pluralist account must therefore permit different standards of value as well as different estimates of consequences. Learning can reveal suitable lives for particular people, previously overlooked tradeoffs, or acceptable terms of cooperation across differences. Those achievements do not require one common destination.

Survival is also ambiguous evidence. An arrangement may endure because it serves people well, because alternatives are unavailable, or because leaving is prohibitively costly. Popularity and persistence deserve investigation; neither can substitute for an account of agency and justification.

Experience can change the evaluator as well as the evaluation. A person may enter an unfamiliar practice with one understanding of what matters and leave with another. That possibility belongs within Millian development. It also means that a fixed measure cannot exhaust an experiment’s significance. We need accounts from participants, opportunities to question those accounts, and room for disagreement about their interpretation.

Experimentation needs a stable platform

An experimental society requires enough continuity for people to undertake meaningful commitments. Someone considering a new career, enterprise, or community needs time and some capacity to survive disappointment. When background conditions are already precarious, the cost of trying something different can become prohibitive.

Maffettone and Muldoon make a related argument about systemic risk. Excessive background uncertainty can undermine planning and the pursuit of a conception of the good; suppressing risk too far can restrict innovation and choice. The institutional question concerns the conditions of agency. [10]

This suggests a practical division of labor. Durable protections support agency and trust. Particular practices remain open to revision. Institutions coordinate consequences that cross local boundaries, while preserving discretion where variation is feasible. Authority should be matched to the reach of the effects, rather than assigned a universally local or central form.

My thesis proposes institutional supports as cognitive prosthetics: public records, independent scrutiny, accessible explanation, and avenues of contestation. Such arrangements must themselves be transparent, contestable, polycentric, constitutionally authorized, and revisable. Otherwise the machinery intended to help citizens judge becomes a machinery for judging on their behalf.

This matters especially where consequences escape the original group. An arrangement can be voluntary for its participants while imposing burdens on neighbors or people downstream. Shared infrastructure can connect otherwise separate attempts. The boundaries of consent, evidence, and authority may therefore differ. Preserving local experimentation requires a workable account of those relationships, including opportunities for affected outsiders to be heard.

The objections are part of the position

Does this turn liberty into permission granted by evaluators? It should do the opposite. Individual agency retains value even when its informational contribution is modest or unknown. The stronger evaluative burden falls on institutions claiming authority over others, not on people whose lives depart from convention. A central board empowered to certify acceptable experiments would defeat much of the purpose.

Does change make all knowledge temporary and unreliable? No. Some regularities endure; some practices work across widely different circumstances; some adaptations preserve what matters despite substantial change. The claim is about warranted application. It asks us to identify which differences matter, rather than to assume that every difference invalidates experience.

Does demanding conditions suppress discovery before it starts? Exploration often reveals questions nobody knew to ask. Its value cannot be restricted to meeting a fixed initial specification. I distinguish exploratory experience from stronger claims about established benefit or general adoption. An unexpected finding is a reason to investigate; uncertainty need not prohibit an attempt.

Why trust institutions to solve a problem caused partly by institutions? There is no general entitlement to that trust. Institutions can conceal errors, concentrate authority, or reward conformity. Their value here is conditional on practical opportunities for scrutiny and correction. Decentralization can distribute mistakes as well as discovery; it needs to earn its claims under the same discipline.

Where this position is most vulnerable

I have not established that contemporary societies generally change faster than they can learn. That claim needs evidence by domain, timescale, and institutional setting. Modern tools can improve feedback and coordination as well as accelerate change. The balance must be investigated.

Neither my workshop paper nor my thesis supplies a measured law linking diversity, update speed, and institutional performance. I now treat the earlier paper’s ratio of diversity to update lag as a heuristic, not an established relationship. More diversity can bring irrelevant variation; faster updating can amplify noise. The institutional argument also does not depend on accepting a particular model of the neocortex.

A useful test would compare arrangements under similar information and resource constraints while varying environmental change, feedback delay, imitation, spillovers, and exit costs. Outcomes should include people’s ability to challenge and leave, the distribution of harm, the preservation of alternatives, and recovery after change. A single aggregate score would conceal part of the question.

The explanatory claim would weaken if supposedly enabling conditions made little difference across relevant cases, or if societies reliably obtained and transferred useful lessons without them. Particular prescriptions should be revised where they reduce agency, prevent worthwhile attempts, or perform worse than available alternatives. The standard applies to this proposal as well.

Keep the possibility of learning open

Most of the ingredients are inherited. Mill supplies individuality and experimentation. Muldoon makes discovery and revision central to political philosophy. My work adds a concern about the feasibility of tracking changing conditions and the institutions needed to support that task. The proposed contribution is their assembly into explicit conditions that can be challenged and studied.

On a dancing landscape, the liberal achievement is a continuing capacity: people can try different lives, understand more of their consequences, recognize when circumstances have changed, and revise the arrangements that govern them. Some experiments will disappoint. Some older practices will remain valuable. Some discoveries will make coexistence easier without producing agreement about the best life.

A society should be judged in part by whether its members retain those possibilities. Freedom gives us room to try. The institutions around it should help ensure that experience leaves us better able to judge what to try next.

Lineage & sources

  1. John Stuart Mill. On Liberty.1859, especially chapter III. Individuality, experiments in living, and the interpretation of experience.Return to the argument ↑
  2. Ryan Muldoon. Expanding the Justificatory Framework of Mill’s Experiments in Living.Utilitas 27(2), 2015, pp. 179–194. Discovery, changing circumstances, and the extension to political justification. Author-uploaded full text.Return to the argument ↑
  3. Ryan Muldoon. Social Contract Theory for a Diverse World: Beyond Tolerance.2016. Perspectives, bargaining, and repeated revision; linked author overview.Return to the argument ↑
  4. Alex Horovitz. The Invisible Limits of Public Reason: Stability, Tracking, and the Political Epistemology of Justification.MA thesis, San Francisco State University, Spring 2026; especially printed pp. 8–13 and 17–20. Author manuscript.Return to the argument ↑
  5. Alex Horovitz. Why We Can’t See the Landscape Dance: Institutional Design Within the Limits of Human Cognition.PPE At the Frontier workshop paper, University at Buffalo, 18 June 2025; especially pp. 4–7 and 10–12. Author manuscript.Return to the argument ↑
  6. Alex Horovitz. True As Of.v0.1, 12 October 2026. Regime-bounded warrant and the conditions of responsible assertion.Return to the argument ↑
  7. Keith Hankins, Ryan Muldoon, and Alexander Schaefer. Does (mis)communication mitigate the upshot of diversity?.PLOS ONE 18(3), 2023, e0283248. A model of communication and the conditions of diversity’s benefits.Return to the argument ↑
  8. Ethan Bernstein, Jesse Shore, and David Lazer. How intermittent breaks in interaction improve collective intelligence.PNAS 115(35), 2018, pp. 8734–8739. Laboratory evidence about the timing of interaction in a problem-solving task.Return to the argument ↑
  9. Gerald Gaus and Ryan Muldoon. Diversity, Discovery, and Justification / Response to Gerald Gaus.Syndicate symposium, published 12 April 2022. Disagreement about evaluative landscapes and the relationship between discovery and justification.Return to the argument ↑
  10. Pietro Maffettone and Ryan Muldoon. On the Normative Relevance of Systemic Risk.Etica & Politica / Ethics & Politics XXII(1), 2020, pp. 511–530. Background risk, planning, and agency.Return to the argument ↑

The intellectual debts are part of the argument. The interpretation and proposed extensions here are my own; citing another author does not imply their endorsement. The thesis and workshop paper are cited by their printed page numbers.

Revision log

Keep learning possible.

Room to try. Time to understand. The standing to question—and the freedom to revise.