Planning is one of the most computationally demanding things the human mind does. To decide on a course of action, people must mentally simulate future states of the world, evaluate the consequences of each possible move, and compare the expected values of different sequences of choices. In a perfectly predictable environment, this simulation can be carried out with confidence: if you take action A, outcome B will follow. But the real world is stochastic. Outcomes are corrupted by noise, the rules governing them shift over time, and sometimes our own actions fail to produce the effects we intend. Each of these forms of randomness, the authors note, makes exact planning exponentially harder, because the decision-maker must average over many possible futures rather than tracing a single one. The theoretical optimal solution to such problems quickly becomes intractable, which raises a fundamental question that has lingered in cognitive science: how do people balance the cognitive cost of planning against its potential benefits when the environment is uncertain?

To answer this question, the team designed a laboratory planning task in which participants encountered one of three distinct forms of stochasticity that are commonly found in real-world environments. The first was reliability, meaning the degree to which outcomes consistently reflect the underlying value of an action. The second was volatility, meaning the rate at which the environment’s structure changes over time. The third was controllability, meaning the extent to which a person’s own actions reliably determine what happens next. These three manipulations capture, in simplified form, the kinds of unpredictability people face when navigating unreliable transportation systems, volatile financial markets, or social situations where other people’s behavior only partially responds to our own. By varying each form of stochasticity independently, the researchers could ask whether different types of randomness affect planning in different ways, or whether the brain responds to all of them with a common strategy.

The central behavioral measure was deceptively simple: the time participants took to make their first choice in each planning episode. First-choice response times are widely used as a proxy for planning depth, because a longer deliberation typically reflects more mental simulation of future possibilities. Across all three manipulations, the researchers found a robust and consistent pattern. As stochasticity increased, people reduced their planning effort, as indexed by faster first-choice response times. In other words, whether the randomness came from unreliable outcomes, a shifting environment, or diminished control over events, the response was the same: participants spent less time deliberating. This consistency across three conceptually distinct sources of uncertainty is one of the study’s most important contributions, suggesting that the brain may treat different forms of environmental randomness through a common computational lens when deciding how much to think.

To understand what was happening inside the mind during these decisions, the team went beyond response times and built a family of computational cognitive models designed to account for participants’ choices. This modeling approach is a hallmark of the Ma laboratory’s work on resource-rational cognition, the idea that the brain optimizes not for perfect performance but for the best performance achievable given limited time, energy, and neural machinery. The models allowed the researchers to distinguish between competing hypotheses about how people cope with stochasticity. One possibility is that people perform something close to the correct calculation, computing expected values by averaging over possible outcomes weighted by their probabilities. Another possibility is that people adopt a much simpler strategy, one that sacrifices accuracy in exchange for a dramatic reduction in mental effort.

The modeling results pointed decisively toward the simpler strategy. Rather than calculating expected values optimally, people behaved as if the world were deterministic, a phenomenon the authors call determinizing. Under this scheme, the planner ignores the probability distribution over outcomes and plans as though each action leads to a single, certain result. Determinizing is computationally cheap: it converts a difficult problem of averaging over many futures into an easy problem of following one imagined path. The cost, of course, is accuracy, because in a genuinely stochastic world the single imagined path will sometimes be wrong. But the benefit is a substantial saving of cognitive effort, and the study suggests that when randomness rises, the brain judges that saving to be worth the loss in precision. This finding connects to a broader theme in recent cognitive science, in which people construct simplified mental representations to make planning tractable, rather than brute-forcing optimal solutions their neural hardware cannot afford.

Within this determinizing framework, the researchers identified a second, subtler signature of reduced effort. As stochasticity increased, people decreased their sensitivity to values, a pattern the authors describe as policy compression. In a fully sensitive policy, small differences in the value of different options translate into large differences in choice behavior, reflecting careful, fine-grained evaluation. When sensitivity to values decreases, choices become less discriminating, as though the planner is working from a coarser, more compressed summary of the decision landscape. Policy compression is a form of lossy compression of a behavioral policy: the brain retains the broad shape of what is worth doing but discards the fine distinctions that would require more computation to resolve. Consistent with the response time findings, this compression deepened as environmental randomness grew, providing converging evidence at both the level of deliberation time and the level of choice structure that planning effort was being actively withdrawn.

The study also carries a methodological warning for the field. The authors emphasize that their results reveal the limitations of studying stochasticity solely through single-shot decisions, the brief one-off gambles that dominate much of the decision-making literature. In a single-shot choice, the computational burden of planning is minimal, and the effects of stochasticity on effort may be invisible. It is only in tasks that require genuine multi-step planning, where the decision-maker must simulate sequences of future actions, that the withdrawal of effort under uncertainty becomes measurable. This suggests that conclusions about how humans handle uncertainty drawn from simple gambling paradigms may not generalize to the richer, sequential decisions that fill everyday life, from choosing a commute route to planning a career move. The planning task developed here offers a template for probing those richer decisions under controlled variation in environmental structure.

The broader implications of the work extend beyond the laboratory. If people systematically reduce planning effort as their environments become less reliable, more volatile, or less controllable, then the cognitive consequences of living in unstable conditions may be more nuanced than commonly assumed. It is tempting to interpret poor planning under uncertainty as a failure of attention or motivation, but this study reframes it as a potentially rational allocation of a scarce cognitive resource: when the expected payoff of careful planning falls because randomness will erode the value of any plan anyway, the brain sensibly stops paying for it. At the same time, the finding raises questions about contexts where reduced deliberation is harmful, such as volatile but high-stakes environments where careful planning remains valuable despite the noise. Understanding when determinizing serves us well and when it leads us astray is a natural next step. The research was funded by the National Science Foundation and the National Institutes of Health, and the authors acknowledge guidance from Marcelo Mattar, Todd Gureckis, and Cristina Savin. As stochasticity shapes everything from economic turbulence to climate variability, this work highlights an often overlooked truth: the mind’s response to an unpredictable world is not to think harder, but to think less.

Subject of Research: How environmental stochasticity affects human planning effort and strategy

Article Title: Environmental stochasticity reduces human planning effort

Article References: Lei, J., Olieslagers, J., Arfaei, N., Lin, D. X., & Ma, W. J. (2026). Environmental stochasticity reduces human planning effort. Nature Communications. https://doi.org/10.1038/s41467-026-78023-9

Image Credits: AI Generated

DOI: 10.1038/s41467-026-78023-9

Keywords: planning, stochasticity, decision-making, cognitive effort, determinizing, policy compression, computational modeling, response times, uncertainty, Nature Communications, resource rationality, human behavior