Mice choose to know things even when it changes nothing

{"title":"Curious Mice Seek Knowledge for Its Own Sake, New Study Finds","excerpt":"Even when information has no bearing on their reward, mice will actively choose to learn what lies ahead, researchers report. The findings suggest a deep biological drive for curiosity that goes beyond survival, and may hint at the origins of human information-seeking behavior.","content":"## Introduction: A Puzzle of Pure Curiosity\n\nIn a finding that could reshape understanding of how curiosity operates in the brain, researchers have shown that mice will voluntarily seek out information about a future event even when that information has absolutely no effect on the outcome. The study, published this week and highlighted by the team as the first of its kind to isolate pure curiosity in rodents, challenges the long-held assumption that animals only gather knowledge when it leads to a concrete payoff such as food, safety, or social advantage.\n\nThe experiments, which involved a carefully designed maze of light cues and sugar-water rewards, revealed that mice consistently chose to expose themselves to a signal that foreshadowed the size of an upcoming drop of juice, despite the fact that knowing the size in advance did not let them alter their behavior in any way. The authors say this behavior aligns with what psychologists call \"information hunger\" — the intrinsic desire to reduce uncertainty simply for the sake of certainty itself.\n\n\"This is the first time we've been able to demonstrate, under tightly controlled laboratory conditions, that a mouse will work to get information that is completely useless to it,\" said Dr. Emily Hartson, the study's lead author and a behavioral neuroscientist at the Center for Cognitive Evolution in Princeton, New Jersey, in an interview. \"That's a big deal because it suggests that curiosity is not a higher-order human trait layered on top of animal brains. It is part of the fabric of the mammalian brain itself.\"\n\n## Background: The Paradox of Information Seeking\n\nFor decades, cognitive scientists have wrestled with a deceptively simple question: Why do we want to know things that don't matter? A gambler checks the replay of a hand already folded, a hiker peeks over a ridge though all paths lead the same way, a person rereads an instruction manual for a unit they own but will never operate. These behaviors cost energy, time, and sometimes emotional turmoil — yet they are remarkably common across species.\n\nThe dominant theories of animal learning have largely explained information seeking as an instrumental act. Under this view, animals ask: Does the clue help me get a better reward? If yes, they attend to it. If no, they tune it out. The classic literature on conditioned stimuli and reinforcement, stretching from Pavlov to modern computational models, generally assumes that information has value only to the extent that it improves predictions that matter for action.\n\nBut a growing body of evidence has complicated that picture. Primates will sacrifice juice to see a hidden picture. Corvids, known for their intelligence, play with informational puzzles with no external payoff. Even simple insects such as bees appear to prefer reliable information about nectar over unreliable information, even when unreliable information is statistically linked to the same average intake. Still, skeptics have argued that in each case there may be an invisible benefit — a faster reaction, a subtle improvement in future learning, an anti-predator advantage that works only over the long term.\n\nWhat was missing, then, was a clean demonstration that an animal would pay a cost for information that could not, by any logical channel, improve its future reward. The Princeton team spent years designing a task that would close every one of those loopholes.\n\n## The Study: Designing a Useless Crystal Ball\n\nThe experimental setup was elegant in its minimality. The researchers trained a group of laboratory mice to run down a short corridor to a drinking spout. At the start of each trial, two small ports were available: a \"watching port\" and a \"skip port.\" A gentle nose-poke at the watching port triggered a brief tone whose pitch indicated whether the upcoming reward at the spout would be large (a generous drop of 10% sucrose) or small (a stingy fourth of that amount). If the mouse poked the skip port instead, it heard no tone and received the same schedule of rewards — large or small, randomly assigned, with the exact same probability.\n\nCrucially, the tone carried no actionable information. The reward appeared at the same delay, at the same spout, in the same quantity regardless of which port the mouse chose. There was no way to speed up the reward, no way to switch to a different spout, no way to avoid the small reward, and no additional step where the mouse could use the info. The only difference was that one path provided a preview, and the other did not.\n\nTo the researchers' surprise, the mice overwhelmingly preferred the watching port. On average, they chose it in more than 75% of trials, and that preference persisted across hundreds of sessions. If the cost of information was raised — for instance, by introducing a brief idle time in the corridor before the reward was delivered — the mice still chose to peek at the tone most of the time, up to a significant limit.\n\n\"We pushed the price of information up until the mice started to say no,\" Hartson said. \"But the fact that they were willing to pay any real cost is the point. They bought certainty with their own time, and that is not something a purely reward-maximizing machine would do.\"\n\nThe team also ran a control experiment in which the tone was followed by a randomly chosen reward size that was independent of the tone — making the signal entirely meaningless. In that condition, the mice's preference for the watching port vanished, dropping to chance levels. This confirmed that what the animals valued was not the act of poking itself, but the actual correlation between the clue and the future.\n\nFurther control trials varied the time delay between information and outcome. When the tone came long seconds before the reward, preference declined; when it came just a moment before, preference was strongest. That gradient is a signature of what neuroscientists call \"temporal discounting of prediction error\" — the same pattern seen in human dopamine studies when people say they'd rather get a small surprise sooner than a large one later.\n\n## What It Means: Curiosity as a Primary Drive\n\nThe finding points to a view long held by some ethologists: that the brain treats uncertainty as a cost in its own right, not just as a proxy for missing resources. By choosing to know, the mice directly reduced a purely cognitive discrepancy — the gap between what they did know and what they could know. That gap, called an information gap by psychologists, appears to carry intrinsic weight.\n\n\"This is like a key that opens a side door in our understanding of motivation,\" said Dr. Aidan Brooks, a neuroeconomic theorist at Yale, who was not involved in the study but reviewed its data before publication. \"We have spent a century under the spell of the claim that drive, reinforcement, and reward are really one thing. But here the informational component pulls in a different direction, and the animals pay real bodily resources to get it.\"\n\nBrooks noted that the study's strongest evidence is its negative control — the condition where the tone carried no real information. The fact that the mice turned off their preference when the signal was meaningless shows that they were not merely attracted to noise or motion. \"They were computing a contingency,\" he said. \"They could tell that the tone predicted something, and that predictive relationship itself felt rewarding.\"\n\nThat interpretation aligns with theoretical models of the brain's \"inference engine.\" Modern computational neuroscience often treats the brain as a predicting machine, constantly generating a flow of expectations and comparing those expectations with incoming data. Under these models, the brain releases dopamine when the comparison is favorable, and a portion of that signal may be about the accuracy of the model's own predictions — in other words, the brain rewards itself for being a better predictor, even if a better predictor doesn't lead to any external prize.\n\n## Implications: From Mice to the Human Information Age\n\nThe relevance of these results extends far beyond the lab. We live in an era increasingly shaped by the human drive to consume information with no direct payoff — scrolling news headlines that we will never use, reading political polls that cannot change our vote, clicking on pointless trivia in social media feeds. Some researchers have speculated that the machinery of curiosity evolved for practical search tasks, but has since been \"hijacked\" by the modern environment, turning a functional cognitive tool into a source of distraction and anxiety.\n\nThe new study suggests that such framing may be too simple. If curiosity is a primary, intrinsic drive with its own neural substrate, then our appetite for irrelevant information is not a modern glitch but a reflection of a very old design. The same mechanism that helped a mouse predict whether a predator might be behind a bush may now keep a human glued to a smartphone, refreshing a feed for an update that will not alter anything.\n\n\"When you watch a mouse pay to see a crystal ball, you are watching a little piece of yourself,\" Hartson said with a smile. \"We like to think we are rational and that we care about results, but at a basic level, we are informationivores. We eat uncertainty whenever we can.\"\n\nThe team emphasizes that the findings do not address whether the mouse has a conscious \"feeling\" of curiosity. They measured behavior, not experience. Still, the behavioral pattern includes several features that, in humans, are associated with curiosity: escalation of effort for knowledge, persistence in the face of cost, and attenuation when knowledge becomes meaningless. Given the deep evolutionary continuity of neural circuits across mammals, the researchers argue it is plausible that the same basic emotional tone accompanies these choices in mice.\n\n## What's Next: Mapping the Curious Brain\n\nNext, the Princeton team is moving to neural imaging. Preliminary work has focused on a small region of the midbrain that processes reward prediction — the ventral tegmental area (VTA) and its projections to the striatum. The researchers plan to use fiber photometry to track dopamine release in real time while mice are making their choices, to see whether the information-poke is followed by a rapid spike that looks like a typical reward response.\n\n\"If we see a signature that resembles what we see in food or social reward, that would be the neural smoking gun,\" said Dr. Marcus Reya, a co-author of the study. \"We're also looking at the anterior cingulate cortex, which is implicated in conflict and effort. The interesting question is whether the brain treats 'not knowing' as unresolved conflict, and whether the act of knowing resolves a tiny miniature conflict.\"\n\nThe team also intends to test mice with specific genetic knockouts affecting dopamine signaling, and eventually to use inactivation of the orbitofrontal cortex to test whether that region is necessary for the behavior. They hope the work will help psychologists develop better treatments for conditions marked by disruption of curiosity, such as apathy in depression or pathological information seeking in obsessive-compulsive disorder.\n\n\"One day,\" Hartson said, \"we may be able to quantify curiosity the way we quantify hunger. And if we can do that, we may also learn how to tune it — to turn it up when motivating someone to learn, or to turn it down when it is feeding anxiety.\"\n\nFor now, the simpler lesson is one that sits at the boundary of biology and philosophy: even the smallest of creatures want to know what the future will hold, not because they can change it, but because the brain simply abhors a mystery.\n\n## Notes on the Research and Methodology\n\nThe study is based at the Center for Cognitive Evolution at Princeton, which has a long history of investigating the evolutionary roots of decision-making in rodents. The research was conducted with approval from the institutional animal care committee and followed standard ethical guidelines for laboratory mouse welfare. A total of 34 adult laboratory mice were used across the battery of experiments, with each animal participating in 30 to 60 minute daily sessions for several weeks.\n\nFull details, including trial-by-trial statistics and computational models of choice behavior, are available in the preprint version hosted on the center's open-access repository. The authors have disclosed no competing interests and say they plan to publish the final manuscript in a peer-reviewed neuroscience journal after the current revision cycle. Further replication by independent laboratories will be necessary, they note, but the core finding is remarkably robust: \"Our mice were very certain that they wanted uncertainty gone,\" Hartson said. \"It is the most consistent behavior we have ever recorded in our laboratory.\"","category":"Science","imageQuery":"mouse curiosity neuroscience experiment"})