Prof. Adam Sanborn
People’s internal representations of natural categories play a crucial role in explaining and predicting how people perceive, learn, and interact with the world. One of the most powerful methods for estimating these representations is Markov Chain Monte Carlo with People (MCMCP) which uses pairwise decisions to sample from very complex category representations. Unfortunately, MCMCP requires a large number of trials to converge, particularly for high-dimensional stimuli such as faces. To address this shortcoming, we integrate a deep generative model, specifically a Variational Auto-Encoder (VAE), into MCMCP, which reduces the dimensionality of the search space and accelerates convergence by using VAE’s implicit knowledge of natural categories. VAE provides MCMCP with a compact and informative representation space via a non-linear encoder, and then focuses human decisions in areas of the representation where the VAE believes the category to be. Otherwise, MCMCP would search in a highly sparse representation space aimlessly until reach its target areas with greater gradients, which is typically lengthy. To test this approach, we ran a new experiment applying VAE-guided MCMCP to recovering people’s representations of happy and sad faces. While past applications of MCMCP to facial affect categories have required chaining across participants, consuming thousands of pairwise decisions before obtaining representative estimates of the means of the two categories, VAE-guided MCMCP converges on an individual’s representation within a single session of less than 150 trials, making MCMCP much more feasible. The study not only provides a method that enables MCMCP to uncover human representations of natural categories more efficiently at individual level, but also provides an innovative and generalizable framework that uses deep neural networks to enhance research into human internal representations.
Prof. Han van der Maas
Van der Maas and colleagues (2020) have recently proposed a mathematical model of opinion formation using Ising networks of attitude elements in order to study polarization and attitude formation A fundamental aspect of this type of network is that it allows for continuous as well as discrete behavior; attention/involvement polarizes attitudes. When individuals are not highly involved in a topic, new information gradually shifts their opinions. However, when individuals are highly involved, their opinions are discrete and require more information to change. This implies that to change someone's opinion, we need to manipulate information without increasing attention. However, explicit attitude measurements can unintentionally increase attention (mere thought effect; Tesser 1978), which has important implications for standard experimental designs. Pretest questionnaires can "freeze" the network, making it difficult for interventions to have an effect - this constitutes a new explanation of an effect known as pretest sensitization. On the other hand, this "freezing" effect may be beneficial for post-testing as this may preserve the effect of the intervention. We propose a study using an extension of the Solomon four-group design to test these hypotheses and identify the most appropriate experimental designs for testing opinion change. In addition to the four-group design (1. pretest-intervention-posttest, 2. pretest-posttest, 3. intervention-posttest, and 4. posttest), we will include three additional conditions: 5. an early pretest-intervention-posttest condition to test whether pretest-induced involvement fades, 6. an intervention-delayed posttest condition and 7. an intervention-posttest-delayed posttest condition to test the effect of posttest intervention preservation. The polarizing effect of attention can have important implications for attitude change interventions and experimental designs, highlighting the need to carefully consider how to measure attitudes and control for potential confounding factors. This study can further advance our understanding of attitude formation and help develop effective strategies for changing attitudes.
Dr. Henrik Singmann
The belief bias is most often investigated with syllogisms varying on two dimensions, logical validity (valid vs. invalid) and believability (believable vs. unbelievable). Typically, participants can distinguish valid and invalid syllogisms (albeit imperfectly), but are also more likely to rate syllogisms as logically valid if they have a believable versus unbelievable conclusion. Additionally, the ability to distinguish between valid and invalid syllogisms can be reduced when their conclusions are believable compared to when they are unbelievable. However, syllogisms are formal reasoning forms unlike arguments we typically see in everyday or informal reasoning. We investigated the belief bias effect in the context of everyday arguments regarding controversial political topics such as those encountered on (social) media (e.g., ‘abortion should be legal’). Arguments in our study differ in their (informal) argument quality; ‘good’ arguments provide an explanation for their conclusion, whilst ‘bad’ arguments do not provide an explanation and contain a reasoning fallacy (e.g., appeals to authority). Participants rated their beliefs about a series of political claims on a scale from 1 to 7 and rated the strength of ‘good’ and ‘bad’ arguments about these claims on a scale of 1 (extremely bad argument) to 6 (extremely good argument). Participants exhibited the belief bias effect for everyday arguments; they consistently rated good arguments as stronger than bad arguments, but were also biased in rating arguments in line with their beliefs as stronger than arguments that were not. The interaction between the quality of an argument and participants’ beliefs about the claims that argument makes is unclear. If we assume the belief and argument strength rating scales are continuous and the relationship between these variables is linear, we fail to find evidence of this interaction using a linear mixed model. However, if we analyse the data using a signal detection approach after binarising the argument strength ratings we find evidence for an interaction, but in an unexpected direction. The ability to discriminate between good and bad arguments increases with the strength of participants’ beliefs about these arguments. The difference in these results is possibly due to assumptions of a nonlinear relationship between the variables in the latter model and raises questions about the most appropriate way to measure the belief bias in everyday reasoning.
Prof. Adam Sanborn
In a complex and volatile environment, agents must learn to capitalize on regularities in their surroundings to be able to infer unknown quantities and generalize past knowledge. However, what properties are learned when adapting to environmental regularities is still unclear. One possibility is that only a few representative examples are stored in memory (Mozer, Pashler, & Homaei, 2008). Alternatively, one could infer rough summaries of the distribution of instances, such as the expected value over all instances and their variability (Tran, Vul, & Pashler, 2017). Finally, agents might acquire full distributional knowledge about environmental regularities, for instance, that instances are distributed according to a powerlaw distribution (Griffiths & Tenenbaum, 2006). These accounts differ in complexity and, as a result, in their ability to capture complex environmental distributions. Intuitively, an agent should strive to match the environmental statistics as closely as possible, with an ideal observer considering the exact environmental statistics as prior information when performing inferences. However, this intuitive notion has been challenged based on theoretical arguments and simulations, arguing that reproducing the environmental frequencies exactly is not robust to possible environmental changes and amounts to overfitting. Instead of veridical environmental frequencies, ideal observers should favor more entropic prior beliefs (Feldman, 2013). Here, we test this argument empirically by contrasting people's beliefs about everyday statistics with their corresponding environmental frequencies. By adopting a novel elicitation technique, random generation as belief elicitation (León-Villagra et al., 2022), we gain access to people's belief distributions for a set of eight domains with different distributional properties. Participants (N=120) produced random draws from everyday quantities (e.g., cake baking times, movie lengths, life expectancies), uttering each item aloud. We transcribed the sequences of utterances, obtaining a trace of "random" draws. Then, we compared the distributions resulting from these draws and the entropy of these distributions to the corresponding environmental distributions and their entropy. We found a significant effect of the domain on the participants' belief entropy, with six out of eight domains resulting in participants exhibiting higher entropy than the environmental data. Our results provide initial evidence for Feldman's theoretical argument for agents' learning beliefs that are more entropic than their encountered environmental frequencies.
Interpersonal synchrony is associated with stronger interpersonal affiliation. No matter how well-affiliated people are, interruptions or other transitions in synchrony rebound to occur. One might intuitively expect that transitions in synchrony negatively affect affiliation or liking. Empirical evidence, however, suggests that time periods with interruptions in synchrony may favor affiliation or liking even more than time periods without interruptions in synchrony. This paper introduces an adaptive dynamical system model to explain how persons’ affiliation might benefit from transitions in synchrony over and above mean levels of synchrony. We evaluated the dynamical system model in a series of simulation experiments for two persons with a setup in which a number of scenarios were explored where different (time) episodes occur. The designed adaptive dynamical system model can be used to model the interaction in therapy or counselling sessions. Its dynamics describe not only the emergence of interpersonal synchrony in such sessions and its adaptive effect on affiliation between therapist or counsellor and client, but also regularly occurring transitions of such synchrony and their adaptive effect.