Discrimination in the environment

This meeting revolves around the influential notion of discrimination in the environment. How do organisms, mostly people, make sense of the deluge of sensory information they're confronted with at any given moment of their life? How does this process go together with learning? What makes it possible? The first talk will cover the role of discrimination learning in language learning, conceptualizing this process as error reduction and showing how it makes it possible to model relevant phenomena of language acquisition without invoking innate, domain specific knowledge. The second talk will question the role of emotions in the discrimination process, stressing their potential role and discussing its implications.

The meeting will take place in room V1, Building Q, at Campus Drie Eiken of the University of Antwerp, from 6 to 8 pm.

The abstracts of the two presentations follow:

Giovanni Cassani (CLIPS): Discrimination in Language

Abstract: In my talk, I will discuss recent developments in the computational modelling of language acquisition using a discriminative learning approach, which relies on the learning framework developed by Rescorla and Wagner in the late 60s to account for the phenomena observed in conditioning. The model conceptualises learning as a competitive process that aims to filter out all the irrelevant information to discriminate the systematic relations between events in the environment. The learning agent is constantly using the environment to predict what will happen next, creating a network of associations that magnifies the informative connections and downplays the noisy, spurious ones. This approach has been recently applied to language processing and acquisition with promising success. I will discuss two studies in some more detail, one investigating cross-situational learning, i.e. how children map words to referents in the world, and another analysing how children learn irregular plurals in English. This are interesting case studies that highlight one of the core principles of discriminative learning, i.e. its ability to learn not only from what co-occurs in the environment but also from what fails to co-occur. Finally, this evidence is used to discuss connectionist models at large, with a focus on their theoretical motivations and how they relate to discriminative learning.


Kris Goffin (Centre for Philosophical Psychology): Affectively Gathering Relevant Information

Abstract: What role does an emotion play in a human's representational system? If you presuppose that the representational system is aimed at mirroring reality, then emotions present us with a projected imagination like reality (as some philosophers have argued), or emotions are just really bad at representing reality as it is. If you presuppose 'relevant realism', i.e. the view that because our senses are imperfect, our representational system has evolved into a system that picks out the relevant truth and avoids unimportant errors in order to achieve the organism's goals, then the function of emotions becomes clear. An important function of emotions is to guide the representational system to the relevant information. In this talk, I will discuss what it means to represent something as relevant.