Our central question is: how accurate are laypeople’sstatistical intuitions about probability distributions within thedomain of health? Specifically, can participants produceentire probability distributions for the duration of illnesses?While a large body of decision making research has suggestedthat people use a flawed process to arrive at decisions, weposit that participants may be using an optimal process, butwith flawed information. To this end, we assess accuracy interms of both the mean and form of distributions for bothacute illnesses for which people might have experience, andchronic conditions for which people are less likely to haveexperience. We find that participants can accurately estimatethe mean and form of distributions for acute illnesses.

People’s predictions for real-world events have been shown tobe well-calibrated to the true environmental statistics (e.g.Griffiths and Tenenbaum 2006). Previous work, however, hasfocused on predictions for these events by aggregating acrossobservers, making a single estimate for the total durationgiven a current duration. Here, we focus on assessingpredictions for both the mean and form of distributions in thedomain of illness duration prediction at the individual level.We assess understanding for both acute illnesses for whichpeople might have experience, as well as chronic conditionsfor which people are less likely to have knowledge. Our datasuggests that for common acute illnesses people canaccurately estimate both the mean and form of thedistribution. For less common acute illnesses and chronicillnesses, people have a tendency to overestimate the meanduration, but still accurately predict the distribution form.

People make accurate predictions for many real world events e.g. human life spans (Griffiths & Tenenbaum, 2006).Accurate predictions are particularly important in the domain of health, where illness knowledge directly influences patientoutcomes. To understand how well peoples’ illness expectations were aligned, we asked participants to estimate durationsfor 9 illnesses, and compared their responses to the real-world distributions. We found that for common acute illnesses (e.g.,the cold) people make accurate predictions, whereas for rare chronic illnesses (e.g., COPD) people make comparatively poorpredictions. Further, we found that participants overestimate the prevalence of every illness, especially for those that are morecommon (e.g., the cold). Taken together, these results suggest that people more accurately estimate the duration of commonacute illnesses, but this may cause them to overestimate the prevalence of these illnesses. Results will be discussed in terms ofimplications for both cognition and behavioral health theory.

Various models have been implemented to explain long-term memory (Brady, et al., 2013; Lew, et al., 2015), withsome being derived from studies of visual working memory (Bays, et al. 2009; Zhang & Luck, 2008). The implicit assumptionis that processes and mechanisms of working memory also exist in long-term memory. However, the findings of fidelity andcontributing factors are highly varied (e.g., Persaud & Hemmer, 2014; Schurgin & Flombaum, 2015) To address what happensto memory traces as they transition from visual working into long-term memory and what factors, such as prior knowledgeand guessing, contribute to the “lifespan” of long-term memory, we implemented three models: the standard remember-guessmodel, a three-component remember-guess model, and a Bayesian mixture model and evaluated these models against data froma continuous recall task. The results clarify the time course of fidelity in long-term memory and pinpoints specific factors thatcontribute to memory.

In a noisy but structured world, memory can be improvedby enhancing limited stimulus-specific memory with statisti-cal information about the context. To do this, people have tolearn the statistical structure of their current environment. Wepresent a Sequential Monte Carlo (particle filter) model of howpeople track the statistical properties of the environment acrossmultiple contexts. This model approximates non-parametricBayesian clustering of percepts over time, capturing how peo-ple impute structure in their perceptual experience in order tomore efficiently encode that experience in memory. Each trialis treated as a draw from a context-specific distribution, wherethe number of contexts is unknown (and potentially infinite).The model maintains a finite set of hypotheses about how thepercepts encountered thus far are assigned to contexts, updat-ing these in parallel as each new percept comes in. We applythis model to a recall task where subjects had to recall the posi-tion of dots (Robbins, Hemmer, & Tang, 2014). Unbeknownstto subjects, each dot appeared in one of a few pre-defined re-gions on the screen. Our model captures subjects’ ability tolearn the inventory of contexts, the statistics of dot positionswithin each context, and the statistics of transitions betweencontexts—as reflected in both recall and prediction.

Temporal Binding (TB) is standardly regarded as an implicit measure of the sense of agency (Haggard, 2017). Though the TB effect is robust, an underlying mechanism has not been agreed upon (Hoerl et al., 2020). Here we propose a memory process as an explanation for the observed error in two publicly available datasets. We first replotted the data and found that on average, across both experiments, participants overestimate the length of the shortest timing interval and underestimate the longest interval, a classic regression to the mean pattern. Summary statistics extracted from the data from each experiment were then used as parameters in a simple Bayesian model of memory. Model simulations reproduced the behavioral data for almost all timing intervals and experimental trial-types across two experiments. Adjusting one of the parameters in the model (prior mean for actions) resulted in an improved qualitative fit. We suggest that other more likely sources of error, apart from experienced agency, may account for this result.