1. What are the origins of variability and biases in numerosity estimations?
The origins of variability and biases in numerosity estimations can be attributed to fundamental principles of acquiring information from environmental regularities. Given biological constraints on information acquisition, numerosity estimation emerges from a system that efficiently considers prior knowledge of the environment, current numerosity being evaluated, and the amount of time available to process such information. This unified normative model incorporates time perception during noisy efficient encoding and optimal Bayesian decoding to predict the corresponding posterior distribution of numerosity estimates. The sequential-encoding/Bayesian decoding (SEB) model and the thermodynamically inspired model (TIM) offer different views on bounded rationality, with SEB attributing variability to sensing costs and TIM attributing it to acting costs during response selection. Empirical tests on a large numerosity estimation data set demonstrate that humans can rapidly and efficiently sample numerosity information over time via an efficient noisy encoding and decoding process.
read more
2. What is the main goal in developing the encoding-decoding model?
The main goal in developing the encoding-decoding model is to incorporate various aspects of human cognition with antecedents in the literature, such as Brownian motion during evidence processing over time and logarithmic internal representation of numerical quantities. The model aims to account for key qualitative features of human behavior with minimal degrees of freedom while acknowledging that different numerosity processing mechanisms may be used depending on the task at hand. Future work could investigate whether situations involving explicit numerosity estimation or discrimination rely on similar or distinct encoding strategies and inference processes. The model assumes a log-normal prior, but also allows for the parameters of the prior to be free parameters, resulting in nonuniform distributions. This theory predicts that larger prior distribution ranges should lead to more noisy estimates and poorer discriminability for a given capacity bound. The model also predicts that the crossover point from overestimation to underestimation should change as a function of the numerosity range. Overall, the encoding-decoding model aims to understand optimal models and the neuro-computational mechanisms underlying human behavior while considering cognitive limitations and the processes of encoding, decoding, and inference.
read more
3. What were the three between-participant experiments in the study?
The three between-participant experiments in the study manipulated available stimulus information. Experiment 1 varied the average dot size, experiment 2 controlled surface density, and experiment 3 controlled surface area of the dots. Each experiment had 110 participants and involved different ways of controlling non-numerical properties of the stimuli. Experiment 4 had a fixed time exposure of 200 ms, but varied the display contrast of the dot arrays between grey and pitch black, with Weber contrasts of 10%, 20%, 40%, 80%, and 160%.
read more