Real-World Modeling

Telling a lie is more cognitively demanding than telling the truth. Support for this notion comes, inter alia, from instructed-lying paradigms showing that untruthful responses are slower than truthful responses. However, conventional measures of the cognitive cost of lying are typically collapsed across response categories and ignore error trials and accuracy, focusing only on average latencies of correct truthful and untruthful responses. To overcome the limitations of conventional approaches and disentangle the mechanisms underlying response behavior in instructed-lying paradigms, we propose to analyze data with a drift diffusion model. The diffusion model considers the full response-time distributions for both correct and error responses, thus making use of all available information. Using a Bayesian hierarchical diffusion model to analyze data from a Sheffield Lie Test, we find that the model’s drift-rate parameter provides for a reliable measure of the cognitive cost of lying. Moreover, we find that truth-vs-lie instructions elicit a response bias that may confound conventional measures that fail to account for it. Thus, our results indicate that the diffusion model constitutes a promising means to analyze data from instructed-lying paradigms and that it offers intriguing avenues for future research on the cognitive mechanisms of lying.

This is an in-person presentation on July 21, 2024 (11:40 ~ 12:00 CEST).

Many learning scenarios involve free practice where learners have the freedom to initiate and stop practice whenever they want (e.g., hobbies and Massive Open Online Courses (MOOCs)). However, a major concern in free practice learning are high dropout rates. Inspired from the literature on learning curves, forgetting curves and motivation-achievement cycles, we propose the Reciprocal-Practice-Success (RPS) model of learning ‘in the wild’. We discuss the rationale behind each component of our model where Success and Practice form a mutually reinforcing positive feedback loop. Through simulations, we show how long term learning outcome is sensitive to the shape of the learning curve; with S-shaped learning curves leading to either expertise or dropout. We also provide a dynamical systems approximation for the RPS model which has a similar qualitative behaviour. Through a bifurcation analysis of two controllable learning parameters - minimum practice rate and success sensitivity, we show what interventions can work to prevent quitting. Next, we show the qualitative results are robust to different forms of the forgetting curve and more realistic extensions to the basic model. We end with a discussion of how our model complements different theories of motivation and self-regulation, with some proposals to reduce quitting in free practice.

This is an in-person presentation on July 21, 2024 (12:00 ~ 12:20 CEST).