Signal Detection Theory

Understanding and predicting the relevant risky choices of modern life is a key goal of behavioral research. However, how do the choices that researchers focus on align with real-life choices that people make? And what are the psychological underpinnings that might explain any differences between the two perspectives? Using a multi-method approach we compiled two comprehensive inventories totaling 165 risky choices, representative of both research and layperson perspectives. Based on cosine similarities of the choices' semantic embeddings and signal detection theory, we identify which choices the two perspectives have in common and where they diverge, and we map their overlap and discrepancies. Moreover, by leveraging the semantic content of choices, we evaluate their relevance to significant real-world risks, as identified by the World Economic Forum and the Global Burden of Disease study. Finally, to better understand why the choices in the research and layperson perspectives are (dis)similar, we examine the role of seven classes of psychological mechanisms (i.e., motivation, experience, affect, cognitive resources, social influence, contextual constraints, and evaluation of choice attributes) in making these choices using Bayesian mixed effects models. In sum, our approach illustrates how well attuned the status quo in behavioral research is to the actual choices and concerns of people. We thereby aim to refine the focus of future studies both in terms of the choices as well as the psychological mechanisms investigated.

This is an in-person presentation on July 21, 2024 (15:20 ~ 15:40 CEST).

A longstanding debate in memory research revolves around the question whether recognition memory judgements are best conceptualized as resulting from the direct comparison of a latent memory signal with a response criterion or through a mediation of memory signals by a small number of latent states. These perspectives have commonly been represented by signal detection theory (SDT) and the two-high-threshold model (2HTM), respectively. Kellen and Klauer (2014) showed that common SDT models and the 2HTM make conflicting predictions in a ranking paradigm and implemented a critical test on that basis; their results were in line with the predictions made by the SDT models and contradicted the 2HTM. However, this conclusion was recently called into question by Malejka and colleagues (2022) who proposed that recognition decisions involving multiple stimuli are based on a contrast mechanism. They argued that if the detection probability for any stimulus in a given set is determined by comparing the memory strength of the stimulus in question with the memory strength of the other stimuli in the set, the 2HTM is able to account for the results of Kellen and Klauer (2014). In order to assess whether a 2HTM that incorporates such a contrast mechanisms is empirically adequate, we directly tested one of its key predictions using a ranking paradigm and found evidence against the model. By contrast, our results align well with the predictions made by SDT. We discuss implications of our results for models of single- and multiple-item recognition memory.

This is an in-person presentation on July 21, 2024 (15:40 ~ 16:00 CEST).

“Better safe than sorry” summarises the principle that strong responses to rare events may be appropriate if the outcome is important enough. The calculation of optimal bias in signal detection theory (McMillan & Creelman, 1991) formalises this idea and is often used to explain the evolution of response biases. However, signal detection theory assumes that payoffs are stable over time. Trimmer et al. (2017) showed that when payoffs change across trials, the optimal bias may have the opposite sign compared to the case of invariant payoffs. Here, we show that the same can happen when payoffs change within trials. We argue that such changes over time are plausible for evolutionary scenarios to which signal detection theory has been applied, and that conclusions drawn from signal detection theory are not always valid when the assumption of time-invariant payoffs is violated.

This is an in-person presentation on July 21, 2024 (16:00 ~ 16:20 CEST).

To make decisions that lead to favorable outcomes, animals need to take into account both their perceptual uncertainty as well as uncertainty about the outcome associated with their actions. There is a long tradition of research investigating how the outcome structure of a task affects animals’ response behavior. The relation between the two has been described by the matching law and its generalizations for tasks with and without perceptual uncertainty. The influence of perceptual uncertainty on decision behavior is often modeled with signal detection theory, which posits that a decision criterion is placed on an internal evidence axis. Where this criterion is placed and how it is updated based on feedback are open questions. Various criterion learning models have been proposed; however, their steady-state behavior across different experimental conditions is not consistent with the aforementioned laws. Our goal is to integrate these approaches to gain a better understanding of the mechanisms by which reinforcements and perceptual uncertainty act jointly to shape behavior. To do so, we first draw an explicit connection between the research on matching laws and signal detection theory by deriving the criterion position that leads to behavior aligned with those laws. Then we develop a learning model that updates the decision criterion trial by trial to learn this criterion position. Our model fits data from a previous experiment well and generates behavior in simulations that is in line with matching laws for perceptual tasks and similar to the subjects’ behavior in the experiment.

This is an in-person presentation on July 21, 2024 (16:20 ~ 16:40 CEST).