Shin and DuBrow propose that a key principle driving event segmentation relates to causal analyses: specifically, that experiences that are attributed as having the same underlying cause are grouped together into an event. This offers an alternative to accounts of segmentation based on prediction error.

When solving problems, like making predictions or choices, people often “sample” possibilities into mind. Here, we consider whether there is structure to the kinds of thoughts people sample by default—that is, without an explicit goal. Across three experiments we found that what comes to mind by default are samples from a probability distribution that combines what people think is likely and what they think is good. Experiment 1 found that the first quantities that come to mind for everyday behaviors and (...) events are quantities that combine what is average and ideal. Experiment 2 found, in a manipulated context, that the distribution of numbers that come to mind resemble the mathematical product of the presented statistical distribution and a (softmax-transformed) prescriptive distribution. Experiment 3 replicated these findings in a visual domain. These results provide insight into the process generating people’s conscious thoughts and invite new questions about the value of thinking about things that are both likely and good. (shrink)

Much of our thinking focuses on deciding what to do in situations where the space of possible options is too large to evaluate exhaustively. Previous work has found that people do this by learning the general value of different behaviors, and prioritizing thinking about high-value options in new situations. Is this good-action bias always the best strategy, or can thinking about low-value options sometimes become more beneficial? Can people adapt their thinking accordingly based on the situation? And how do we (...) know what to think about in novel events? Here, we developed a block-puzzle paradigm that enabled us to measure people's thinking plans and compare them to a computational model of rational thought. We used two distinct response methods to explore what people think about—a self-report method, in which we asked people explicitly to report what they thought about, and an implicit response time method, in which we used people's decision-making times to reveal what they thought about. Our results suggest that people can quickly estimate the apparent value of different options and use this to decide what to think about. Critically, we find that people can flexibly prioritize whether to think about high-value options (Experiments 1 and 2) or low-value options (Experiments 3, 4, and 5), depending on the problem. Through computational modeling, we show that these thinking strategies are broadly rational, enabling people to maximize the value of long-term decisions. Our results suggest that thinking plans are flexible: What we think about depends on the structure of the problems we are trying to solve. (shrink)

Topics in Cognitive Science, Volume 14, Issue 3, Page 508-527, July 2022.

Recent debates over adults' theory of mind use have been fueled by surprising failures of perspective-taking in communication, suggesting that perspective-taking may be relatively effortful. Yet adults routinely engage in effortful processes when needed. How, then, should speakers and listeners allocate their resources to achieve successful communication? We begin with the observation that the shared goal of communication induces a natural division of labor: The resources one agent chooses to allocate toward perspective-taking should depend on their expectations about the other's (...) allocation. We formalize this idea in a resource-rational model augmenting recent probabilistic weighting accounts with a mechanism for (costly) control over the degree of perspective-taking. In a series of simulations, we first derive an intermediate degree of perspective weighting as an optimal trade-off between expected costs and benefits of perspective-taking. We then present two behavioral experiments testing novel predictions of our model. In Experiment 1, we manipulated the presence or absence of occlusions in a director–matcher task. We found that speakers spontaneously modulated the informativeness of their descriptions to account for “known unknowns” in their partner's private view, reflecting a higher degree of speaker perspective-taking than previously acknowledged. In Experiment 2, we then compared the scripted utterances used by confederates in prior work with those produced in interactions with unscripted directors. We found that confederates were systematically less informative than listeners would initially expect given the presence of occlusions, but listeners used violations to adaptively make fewer errors over time. Taken together, our work suggests that people are not simply “mindblind”; they use contextually appropriate expectations to navigate the division of labor with their partner. We discuss how a resource-rational framework may provide a more deeply explanatory foundation for understanding flexible perspective-taking under processing constraints. (shrink)

The dominant practice of AI alignment assumes (1) that preferences are an adequate representation of human values, (2) that human rationality can be understood in terms of maximizing the satisfaction of preferences, and (3) that AI systems should be aligned with the preferences of one or more humans to ensure that they behave safely and in accordance with our values. Whether implicitly followed or explicitly endorsed, these commitments constitute what we term apreferentistapproach to AI alignment. In this paper, we characterize (...) and challenge the preferentist approach, describing conceptual and technical alternatives that are ripe for further research. We first survey the limits of rational choice theory as a descriptive model, explaining how preferences fail to capture the thick semantic content of human values, and how utility representations neglect the possible incommensurability of those values. We then critique the normativity of expected utility theory (EUT) for humans and AI, drawing upon arguments showing how rational agents need not comply with EUT, while highlighting how EUT is silent on which preferences are normatively acceptable. Finally, we argue that these limitations motivate a reframing of the targets of AI alignment: Instead of alignment with the preferences of a human user, developer, or humanity-writ-large, AI systems should be aligned with normative standards appropriate to their social roles, such as the role of a general-purpose assistant. Furthermore, these standards should be negotiated and agreed upon by all relevant stakeholders. On this alternative conception of alignment, a multiplicity of AI systems will be able to serve diverse ends, aligned with normative standards that promote mutual benefit and limit harm despite our plural and divergent values. (shrink)

Topics in Cognitive Science, Volume 14, Issue 3, Page 528-549, July 2022.

How much would you pay to play a lottery with an “infinite expected payoff?” In the case of the century old, St. Petersburg Paradox, the answer is that the vast majority of people would only pay a small amount. The authors seek to understand this paradox by providing an explanation consistent with a broad, process‐level model of human decision‐making under risk.

The idea of replication is based on the premise that there are permanent laws to be replicated and verified, and the scientific method is adequate for doing so. Scientific truth, however, is not absolute but relative to time and context, and the method used. Time and context are inextricably interwoven, in that time creates different contexts and contexts (e.g., Christmas Day vs. New Year’s Day) create different experiences of time, rendering psychological phenomena inherently variable. This means that internal and external (...) conditions fluctuate and are different in a replication study vs. the original. Thus, a replication experiment is just another empirical investigation that has no special status in the establishment of scientific truth. It is not the final arbiter of whether or not something exists. In their pursuit of homogeneous external conditions, replications have ignored the homogeneity of internal conditions. There is not a single replication reported in the literature that would have shown participants’ feelings and thoughts—both conscious and nonconscious—to be identical to those of the original participants. Experimental instructions can create varying ratios of conscious over nonconscious processing from one study to another. Ironically, every replication is a failure at the fundamentals of human psychology. While patterns can be discovered, they are not permanent or unchangeable laws of human behavior to be proven by the pinpoint statistical verification through replication. As scientific knowledge in physics is temporary and incomplete, should it be any surprise that science can only provide “temporary winners” for psychological knowledge of human behavior? (shrink)