#ComCo2023 was a blast! Two days packed with exciting research: six keynote talks, eight contributed talks, two poster sessions and countless awesome interactions after the sessions, during the lunch/coffee breaks, at the bar and during our networking and warm up sessions!
Special thanks go to our keynote speakers: Valerio Mante, Asya Achimova, Carlos Zednik, Marieke Woensdregt, Tom Froese and Dora Angelaki. Thank you for coming and sparking so many exciting discussions - your talks were an inspiration.
Nice methodological work testing a velocity-based algorithm which showed accurate classification of eye movements into gazes and saccades, and that ERSPs are better suitable to study behavior in free-viewing experiments as they are less time-sensitive.
Keynote talk by @DrTomFroese from @OISTedu on Making Mind Matter. Lots of food for thought! In this talk the irruption theory was presented, which is a novel approach to conceptualize the enactive account of motivated activity.
In essence, it was propsed that cognitive science has been misguided by focusing on observables that are in principle fully explicable in terms of other observables.
Practically there needs to be a switch of focus and a development of the cognitive science toolbox for identifying those observable changes that are inexplicable in terms of observables.
Discussion included the two experiments conducted using demonstratives to extract a computational model which accounts for the visual search cost of the receiver with softmax rationality parameter.
Models were compared and results showed no difference between the attention-correction model versus baseline. Additionally the talk included strategies for resolving ambiguity, and future work could focus on how humans manage resource demands.
#pragmaticreasoning#languagemodels
Does explainable AI needs cognitive models? last keynote of the day by @CarlosZednik who discusses the advantages and limitations of “top-down” and “bottom-up” explainable strategies.
Importantly both strategies are limited to promote trustworthiness in AI. This begs the question whether we could seek inspiration and adopt an explanatory strategy from the field of cognitive science.
Specifically also the challenges of the cognitive modeling approach including interpretability, scalability, etc. #ExplainableAI#cognitivemodels#AItrust
from neural spiking synchrony (@Vik_Zemliak) to applying computational cognitive models to real-world driving scenarios (Gustav Markkula) to now poster session with some nice ☕️&🍰 #ComCo23
Exciting kick-off to #ComCo23 this morning: Using recurrent neural networks to translate neural dynamics of primates to decision behavior (Valerio Mante), discussions on pragmatic language use with GPT-3 (Polina Tsvilodub),