Date: May 22, 2025
Location: Washington Hilton, 1919 Connecticut Ave. NW, Washington DC
Why do we avoid effort even as we seek it out? This talk begins with the law of least effort, a foundational principle in psychology and economics that states, all else equal, organisms prefer actions requiring less effort. The evidence for this is robust: people avoid cognitively demanding tasks, fatigue quickly, and even prefer pain to effort in some contexts. Yet, paradoxically, we also value effort—gravitating toward challenging goals, deriving meaning from strenuous pursuits, and sometimes pursuing tasks precisely because they require effort. I will explore this tension and offer three possible explanations for why the effort paradox exists. First, that the paradox is illusory: effort's apparent value arises from misattribution of emotions from outcomes rather than effort itself. Second, that effort-seeking is completely explained by reward maximization, consistent with the law of least work. And third, that the paradox is adaptive: valuing effort may help regulate how much we exert ourselves, avoiding both under- and over-exertion. Ultimately, I argue that effort’s dual nature helps explain its central role in human motivation.
We make a great many decisions– who to marry, whether to have kids, where to live – on the basis of how we expect those choices to make us feel in the future. Why then do we sometimes choose the “boring” option, even at the expense of more meaningful alternatives? Across several studies, we find evidence that boredom is an affective signal that flags deficits in meaning or attention that impede successful goal pursuit. Despite this, people often paradoxically choose to pursue boring options at the expense of more meaningful alternatives. Across several field (COVID-19 vaccines, Thanksgiving, 2020 election) and lab (kava ritual, scuba diving) studies, we provide evidence that people routinely underestimate the meaning of future events, and that such errors (in combination with effort avoidance) lead people to prioritize the pursuit of easy meaningless (“boring”) tasks over more effortful but also more meaningful ones.
Boredom presents an essential conundrum for the bored individual, signaling the strong desire to be engaged in something meaningful, while making obvious that there are few options the individual thinks will satisfy. Work has long shown that those high in boredom proneness tend to also be low in self-control. The tool most commonly used to measure self-control, however, focuses more on outcomes than mechanisms. I present data showing that the highly boredom prone do not struggle with impulse control, but rather, suffer from a failure to launch into action. I then tie this claim to recent work from our lab showing that boredom proneness is associated with increased attention to interoceptive signals, coupled with challenges in making sense of those signals. Placing this difficulty within a predictive coding account suggests that the highly boredom prone have faulty predictive expectations for what might count as engaging, which potentially underlies their failure to launch.
Having a longer-term goal with a deadline can fundamentally shape motivation. The closer we get to a deadline, and the more work that remains to be completed, the greater the pressure there is to exert more effort to ensure we complete the goal in-time. However, to-date, there has been no formal computational account for this ‘deadline pressure’, nor an examination of the underlying neural mechanisms. I will present four studies using a novel task that shows effort-based decisions are significantly impacted by deadline pressure. Using computational modelling and ultra-high-field fMRI, I offer an account for how high deadline pressure shifts people to seek (and perhaps value) high effort, and highlight the multiple computations performed across frontal and basal-ganglia regions underlying these dramatic changes in preferences. This work offers a novel account of the neurocomputational mechanisms underlying deadline pressure and how they keep us ‘on track’ to complete our goals.
Flow is a coveted psychological state characterized by deep immersion and engagement in an activity. While its benefits for productivity and health are well-documented, a formal, mechanistic understanding of the flow-generating process remains elusive. In this talk, I will present a solution: a mathematical model of flow's computational substrates—the first of its kind—supported by empirical tests of its core predictions. At the heart of the model lies the concept of mutual information, a fundamental quantity in information theory that quantifies the strength of association between two variables. The central claim is that the mutual information between desired end states and means of attaining them, or I(M;E), gives rise to flow. I will substantiate this claim with behavioral experiments demonstrating that, across multiple activities, increasing I(M;E) increases flow and has important downstream benefits, including enhanced attention, enjoyment, and skilled performance.
Room: Lincoln East
Research on self-regulation has often focused on how people exert control over their thoughts, emotions, and behavior. Relatively less attention has been paid to how or if people manage their motivational states in the service of achieving valued goals. I will discuss a program of research that focuses on the role of people’s beliefs about motivation (i.e., their metamotivational knowledge) in effective goal pursuit. At the heart of this framework is the idea that qualitatively distinct motivational states often involve performance tradeoffs, such that a given motivational state (eagerness) may be relatively beneficial for some tasks (creative brainstorming), but detrimental for others (detailed editing). Thus, effective goal pursuit for both the self and others involves both discernment and flexibility in the regulation of motivation.
At any moment in time, the brain has access to many cognitive strategies available to solve tasks. Because these strategies coexist, we face the problem of metacontrol: how do we decide which strategy to use? And how do we make this arbitration process? I will present several lines of recent research in my lab suggesting that people lean on task structure to minimize metacontrol. Our research combines behavioral paradigms, computational modeling, and neuroimaging to demonstrate that people use their knowledge of task structure to find adaptive behavioral policies. We demonstrate this in various research domains, including cognitive control, planning, and human-AI interactions.
Emotion regulation is important for psychological well-being, yet we know relatively little about why, when, and how hard people try to regulate emotions. To address these questions, in this talk, I will consider effortful emotion regulation as a form of cybernetic control. I will propose that both healthy and unhealthy emotion regulation may be critically informed by the emotion goals people pursue, the perceived discrepancy between such goals and current emotions, emotion regulation goals, and a cost-benefit analysis that determines how much effort people invest and which regulatory behaviors they engage in. I argue that this approach to emotion regulation gives rise to novel research questions. I will highlight some related questions, share empirical research that speaks to them, and discuss potential implications.
I explore the relationships between goals, which have implications for motivation and achievement. Goals are cognitively represented in harmony or conflict. The perception of goal harmony fosters intrinsic motivation, making pursuing long-term interests more appealing. Alternatively, the perception of goal conflict activates self-control to overcome immediate temptations in pursuit of long-term interests. I discuss when perceiving harmony (e.g., between health and taste) enhances motivation and when recognizing conflict (e.g., in intertemporal choices) bolsters motivation. By examining these possibilities, I propose that effective self-regulation often depends on strategically representing goals in harmony or conflict.
Room: Lincoln West
When making decisions, time is of the essence. Deliberate too little and risk unwarranted mistakes. Deliberate too long and risk other, more valuable opportunities. This fundamental speed-accuracy tradeoff underpins normative models of sequential decision-making, such as the drift-diffusion model (DDM), where optimal deliberation times are those that maximize the decision-maker's instantaneous rate of reward. Humans fall short of this normative standard. Why? Here we propose a novel answer: People calibrate their deliberation times to maximize not just their rate of reward, but also their rate of information gain, in accordance with an emerging body of work suggesting that humans and non-human animals have a fundamental drive to acquire information. We present a modified DDM in which choices are made by maximizing a combination of information gain and reward rate. This model captures behavior that has systematically deviated from previous models when a task is difficult, learnable or both. In fact, it offers a normative case for itself when the possibility of learning is taken into account because it leads to more efficient learning. Our proposal suggests that the need for information shapes not only what decision-makers choose, but also when decision-makers choose. If our proposal is correct, then the objective function that governs deliberation time is fundamentally different than previously thought: it considers not just instrumental reward, but also information gain.
Why do people seek to improve themselves? One explanation is that improvement is intrinsically rewarding. This can be formalized in reinforcement learning models by augmenting the reward function with intrinsic rewards (e.g., internally generated improvement signals). In this talk, I propose an alternative explanation: the drive for improvement arises from planning in a state space that includes internal states (e.g., competence). Planning is therefore reflective in the sense that it considers the value of future internal states (e.g., “What could I accomplish in the future if I improve my competence?”). I will present a reflective Markov Decision Process (rMDP) framework that formalizes this idea as a sequential decision problem. I will show that the model captures qualitative patterns of skill development better than a range of alternative models that lack some of its components. Importantly, it explains these patterns without appealing to intrinsic rewards.
Anhedonia is common across mental health disorders and is associated with poor treatment outcomes. We build on reinforcement learning (RL) theories of physiological drive and appraisal to provide a novel mechanistic account of consummatory anhedonia. We reconceptualise hedonic experience in terms of inferred progress towards personally meaningful goals. Through three independent and replicated behavioural studies, we use an RL framework to show that subjective pleasure is best explained by perceived progress towards real-world personal goals in a proximity-dependent manner. We also show that introducing a second orthogonal goal dimension reduces inferred progress and results in reduced hedonic response. Perceived progress towards a goal state elicits subjective pleasure, implying a fundamentally motivational component to hedonic experience. We suggest that consummatory anhedonia may arise from any change which increases perceived goal distance. This accounts qualitatively for the hedonic effects of stress and maladaptive cognition, offering novel views on anhedonia treatment.
Despite the importance of motivation to an understanding of human behavioral variation—and to efforts to modify problematic patterns of behavior—few algorithmically-explicit accounts of motivation are currently available. In our proposed article, we consider the possibility that motivated behavior emerges via interactions between simple processing units, which are commonly modeled via neural networks (NNs). The neural network framework is an approach to understanding how cognition arises from brain processes. It has provided generative accounts of many types of human cognition and has also provided the foundation for modern AI systems, including Large Language Models (LLMs). Despite its successes, the NN framework has not yet been systematically applied to understanding motivated behaviors. In this article, we argue that the time is ripe for just such an application. Our central proposal is that it is useful to conceive of (neural) activation as the common currency of motivated behavior. When modeled in a neural network, the levels of activation that emerge in an output layer can correspond to an agent's behavior and its decisions. We use the NN framework to describe the emergence of the influence of needs, emotions, and goals on behavior. On our account, these processes must be explained—rather than being treated as used to explain other aspects of motivated behavior. The NN framework affords an opportunity to achieve this objective.
Room: Lincoln East
Psychologists have not always appreciated the power of opportunity. Some of our field's most prominent early tests claimed to measure innate capacities but favored the native-born American elite, leading to conclusions that higher SES groups were inherently superior. This oversight shaped policies on immigration, education, and poverty. Here we examine whether common measures of motivation today might exhibit this same bias. To test this, socioeconomically diverse participants completed widely used behavioral scales that measure an individual's altruistic, environmental, and health motivation. Across all studies, low SES participants' scores indicated lower motivation for altruism, environmentalism, and health responsibility than high SES participants. However, we also had participants indicate for each scale item whether they have ever had a reasonable opportunity to perform the action (and if not, why). Across studies, low SES participants reported fewer opportunities to perform actions on each scale. Critically, the performance gap between SES groups was driven by the disparity in opportunity: when we corrected for opportunity, these group differences vanished. These findings indicate that a correspondence bias shaping our field's earliest measures may persist in our measures of motivation today.
The government can address economic inequality through benefits programs that provide aid to lower-income people. However, participation in these programs is consistently low. Considerable research has examined why aid seekers do not apply, but there is limited research on aid-seeking motives. Using mixed methods, we document common practice and predictions for encouraging aid seeking, and evaluate these approaches with eligible recipients. While common practice and predictions endorse the approach of describing how aid can fulfill physiological needs (i.e., hunger), we find that aid seekers are most motivated when aid is framed as advancing interdependent psychological needs (i.e., the need to strengthen community ties). In a large field experiment (n = 4,968), an intervention that described how aid promotes interdependence increased eligible participants' intentions to seek aid and increased their requests for application resources by 10% relative to the control. Aid-related self-efficacy, or the perception that the aid enables recipients to reach their goals, explained these effects. The interdependence treatment was the only condition to shift these consequential outcomes, and the effects persisted six months later. Promoting aid as central to survival may fall short of the benefits of promoting aid as the key to thrive.
Artificial intelligence (AI) is becoming ubiquitous. While many of us already rely on AI as a tool to achieve our goals, AI will—based on personalized algorithms—increasingly be able to set our goals for us. In this project, we investigate the downstream consequences of AI goal-setting for human motivation. In three studies, participants could choose between low and high rewards and complete an equation to earn them. Sometimes, however, this choice was made by a rule-based AI agent that always selected the high reward. The value of the highest reward option varied (20 or 5 cents) across trials, and we quantified performance by speed and accuracy. Study 1, with mixed AI- and self-choice trials, revealed an overall value effect, with higher accuracy for higher rewards, regardless of who made the choice. Study 2, employing separate blocks of AI- and self-choice trials, detected this value effect only for the self-choice condition. In Study 3, we replicated this interaction effect between value and self-choice on accuracy and further tested whether this effect was due to a lack of attention to the reward AI chose. A condition where participants had to confirm the AI's choice with a key press did not eliminate the value effect but revealed lower speed performance for AI (vs self-choice) trials. Together, this suggests that AI goal-setting may reduce the effect of goal value on motivation and possibly even motivation in general.
Our research examines whether "motivation knowledge"—individuals' understanding of how to motivate themselves—predicts successful goal pursuit. Across three studies (n = 900), we identify two groupings of self-regulatory strategies: radical approaches (e.g., all-or-nothing framing, strict rules, immediate change) and gradual approaches (e.g., flexible framing, intuitive rules, incremental change). Additionally, we address two empirical questions: (1) Do radical (vs. gradual) approaches better predict self-regulatory success and (2) Do people accurately predict this relationship? Our findings suggest that while the endorsement of both approaches predicts success for healthy-eating goals, the endorsement of radical strategies is a stronger predictor. We replicate these results in the exercise and finance domain. However, people's beliefs about the effectiveness of these strategies vary by domain. People tend to underestimate the effectiveness of radical strategies for health-related behavior change, but accurately recognize the effectiveness of radical strategies in the financial domain. These results indicate that while people believe gradual (vs. radical) strategies to be most effective for achieving their health goals, successful self-regulators are more likely to endorse radical strategies.
Attendees are invited to enjoy lunch at their leisure from the many nearby restaurants. We've provided some great nearby options below. Attendees are also welcome to join the SSM Executive Committee's annual business meeting for news about the society, its initiatives, journal updates, and more; it will be held in Lincoln West from 1:15 - 2:15 PM.
Room: Lincoln West
Curiosity motivates exploration and supports learning, but curiosity is not always experienced in the face of the unknown. Why does curiosity arise in some situations but not others? In this talk, I will present research investigating the extent to which curiosity is experienced selectively when a promising learning opportunity is present. Broadly, this work shows that curiosity (1) prioritizes immediate learning gains over progress towards our ultimate learning goals, and (2) fails to account for the practical costs of information search. Thus, curiosity is an imperfect guide to learning.
How do people decide what they want to know? In this talk, I will present research demonstrating that people intuitively assess information based on three key factors: (i) its usefulness in guiding action, (ii) its emotional impact, and (iii) its role in improving our understanding of the world. These factors can be distilled into two fundamental components of value: extrinsic (instrumental) value and intrinsic (non-instrumental) value. Our findings reveal that these two components are represented by distinct brain systems: while the traditional reward system (e.g., VMPFC, striatum) selectively encodes intrinsic value, regions involved in perception encode extrinsic value. This pattern holds across different modalities (e.g., visual, linguistic). These insights allow us to predict how information-seeking behavior may be altered in disorders where the reward system is impaired. We test these predictions through a linguistic analysis of participants' web searches, relating search patterns with self-reported psychiatric symptoms. Finally, we introduce a novel tool that leverages language analysis algorithms to evaluate and score information based on its usefulness, emotional valence, and ability to enhance understanding. These scores are displayed alongside web search results, guiding users toward information that best fulfills their instrumental, cognitive and emotional needs.
Converging evidence suggests that animals maximize expected information gains (EIG)—the early resolution of uncertainty about an action or state—but little is known about the mechanisms of this process. I will present neurophysiological evidence that EIG computations are associated with the fronto-parietal network involved in spatial attention control, and describe a neurocomputational model of attention allocation to maximize EIG. The model involves meta-level regulation of spatial attention by an executive circuit and distinguishes between the valuation of an attentional policy (in the executive circuit) and the implementation of the policy (in visual structures). I will show that the model significantly outperforms traditional reinforcement learning approaches in explaining neural activity and provides mechanistic explanations to individual variability in information demand and curiosity, underscoring the value of unifying research on attention and curiosity to shed light on the mechanisms of both functions.
Work on curiosity is closely tied to work on rational learning. Researchers have proposed that we are motivated to explore in the face of information gaps, opportunities for expected information gain, reductions in uncertainty, moderate amounts of surprisal, increased empowerment, or high rates of learning progress. Although these accounts differ in many respects, they are united by the idea that organisms are motivated to build more accurate models of themselves and the world. Juxtaposed against this work (here, quite literally in the parallel session), a different tradition has focused on the motivation to pursue social and identity-based goals; goals that may lead away from rational learning and towards partisan biases and belief polarization. Here I present new work that crosscuts these distinctions. Children’s insight learning is affected by their identification with a social group: Children are faster to overcome functional fixedness when the initial function of a tool is demonstrated by peers of a different gender than peers of the same gender. Unlike motivated reasoning, these inferences are neither a reaction to cognitive dissonance nor self-serving. However, unlike other effects of social context on cognition, neither are these inductive biases likely to help learners draw accurate inferences. One speculation is that the ways that social motivations constrain learners’ inferences contributes to cultural variability and the division of labor.
Room: Lincoln East
Across the social sciences, there have been many models of belief polarization. Most (but not all of these) include an assumption that individuals with substantively different beliefs tend to influence each other less than those with similar beliefs. This said, there are a number of different, plausible mechanisms by which this can happen, and thus a number of different, plausible paths to belief polarization. This talk will discuss these different mechanisms, and draw some upshots for the empirical study of polarization. First, we should not expect that all instances of belief polarization are driven by the same mechanisms, and second real cases of polarization are likely often driven by multiple, interrelated causes.
I will discuss experiments from Celniker & Ditto (2024) testing the impact of partisans’ preferences and prior beliefs on partisan bias. Participants were randomly assigned to make blinded or unblinded evaluations of the same scientific study. The blinded evaluations—judgments of a study’s quality given before knowing whether its results were politically congenial—served as impartial benchmarks against which unblinded participants’ potentially biased evaluations were compared. Modeling the influence of partisans’ preferences and priors across conditions allowed us to jointly test “hot” and “cold” accounts of partisan bias. We found that preferences and priors independently biased unblinded partisans’ evaluations relative to their blinded counterparts’ evaluations. Nevertheless, the observed biases were better explained by the shared influence of preferences and priors, rather than their independent effects. I will end by considering what the functional entanglement of these variables may suggest for studying and preventing partisan biases.
Recent research finds that even entrenched conspiracy theorists will change their minds when an AI debunker presents them with counterevidence that is specific to their unique beliefs. We argue that the effect is strong and durable because the AI provides such strong evidence-based counterarguments. However, if people believe that they are conversing with an AI chatbot that was trained by a political out-group on counterpartisan content, will motivational processes diminish or even eliminate the effect of evidence? Results for this new experiment will be presented.
The opinions people form depend on their motivations. A common portrayal in politics suggests that people’s attitudes reflect directional motivations to align with their party affiliation (e.g., Democrat or Republican). Yet, there is limited work that directly documents the motivational basis of political opinions. Moreover, no scholarship has unraveled whether partisan motivations stem more from adherence to group identity norms or a belief in the informational value of the group’s wisdom. We use an experiment to manipulate feelings of partisan affiliation or motives to rely on advice from the party. The topical focus of the experiment is climate change – a domain of obvious import. We additionally explore the impact of motivations on beliefs, policy positions, and intended climate behaviors, as well as those related to mitigation and adaptation.
Room: Lincoln West
We will announce the winners of the Young Scientist Poster Award, the Early Career Award, and the Distinguished Contribution to Motivation Science Award.
Psychology has been criticized for the weakness of its theories. Nowhere are these concerns more obvious than in the field of motivation. Despite the importance of the concepts, psychology has been slow to develop major theories of motivation and goal pursuit. This may be the case because, as Lewin (1935) lamented, we are stuck in an Aristotelian mode of thinking characterized by a tendency to focus on specific phenomena rather than on the general laws of behavior and to classify concepts based on their value rather than according to the nature of the psychological process involved. In this talk, I am reminding motivational scientists what Lewin advocated for 100 years ago: we need a Galilean approach, with no value concepts and dichotomies, but based on common psychological laws across behaviors and phenomena of interest. Without that, we risk being irrelevant and making little progress.
This talk examines a growing tension in motivational science: the increasing emphasis on methodological rigor alongside a decline in theoretical ambition. While the field has achieved significant advances in experimental design, statistical sophistication, and empirical standards, questions remain about the purpose guiding this progress. The talk introduces the “Quandary of Purpose”—a quiet imbalance between precision and meaning, between data collection and conceptual insight. Theory, when treated as a central and teachable discipline, serves not only to organize knowledge but to propel it forward. Yet many emerging scholars are trained to master methods without ever learning how to build or challenge frameworks. This address calls for a revitalization of theory-building as a foundational scientific skill and advocates for a culture that values both methodological excellence.
Room: Lincoln West
AI researchers who study "reinforcement learning” have developed the idea of an intrinsic reward signal called “empowerment” which maximizes mutual information between actions and their outcomes. Seeking empowerment motivates a search for controllable aspects of the environment, regardless of whether they are directly rewarding. “Empowerment” may be an important bridge between classical Bayesian causal learning and reinforcement learning and may help to characterize causal learning in humans and enable it in machines. Empowerment may also explain distinctive empirical features of children’s learning. In several empirical studies, we systematically test how children, adults and AI agents use cues to empowerment to motivate exploration and learn about their surroundings.
Humans are emotional beings, that much is clear, and emotion seems to interfere with our ability to behave more “rationally”. But the more we look at the brain of humans and nonhuman animals alike, the more it seems that the brain is not organized in terms of clear-cut, separate systems that support “emotion”, “cognition”, “motivation”, and so on. In this talk, I will present evidence that these mental capacities are supported by dynamic and distributed processes that span the entire brain. This organization supports a degree of “computational flexibility” that enables animals to cope successfully with complex and ever-changing environments. An implication of this framework is that brain processes do not respect the boundaries of standard mental terms: there are no separate brain territories for perception, cognition, action, emotion, or motivation.
Most theories of motivation rely on goals, values, and/or feelings (affect/emotion) to explain behavior. However, the nature of each of these variables remains almost entirely unexplained. For instance, how do we select the right goal to have at a given moment? How do we arbitrate between different sources of value, and when and how are these informed by our feelings? I will propose a way out of this entanglement, which centers motivation on a single source of value – affect. Specifically, I will argue that all forms of motivated behavior can result from dynamically adjusting our actions and control states to achieve outcomes that would feel better and avoid states that would feel worse. This Affective Gradient Hypothesis overcomes obstacles to solving motivated behavior by suggesting that goals emerge from affect optimization (rather than being directly selected), and that there are no forms of value beyond affect (e.g., “cold” values).
Location: Mission Dupont | 1606 20th Street NW | ~9 min walk
Wrap up a great day with food, drinks, and conversation at Mission Dupont, featuring an open bar and complimentary appetizers.