[SydPhil] HPS Research Seminar - Dr Julia Haas - MORAL GRIDWORLDS
debbie.castle at sydney.edu.au
Mon Oct 14 13:56:28 AEDT 2019
View this email in your browser<*|ARCHIVE|*>
SCHOOL OF HISTORY AND PHILOSOPHY OF SCIENCE
IN CONJUNCTION WITH THE
SYDNEY CENTRE FOR THE FOUNDATIONS OF SCIENCE
SEMESTER TWO 2019 RESEARCH SEMINAR SERIES
School of Philosophy, Research School of Social Sciences
The Australian National University
The nascent field of moral AI asks how best to ensure that artificially intelligent (AI) systems instantiate and promote human moral norms and values. A growing number of proposals have suggested that the best way to achieve this objective is by developing artificial moral cognition, that is, artificial systems that recognize and respond to situations of moral significance. However, extant approaches involve building sophisticated cognitive architectures into these artificial systems, resulting in intractable frameworks. I present a new technical framework for modeling artificial moral cognition. Specifically, I propose to use reinforcement learning-based environments that teach artificial systems to learn, value, and respond to moral content and contexts. To illustrate the view, I describe a basic “moral gridworld,” designed to model the moral cognitive capacity of fairness, here quantified in terms of the Ultimatum Game. I then discuss a suite of variations that can be introduced into the gridworld to represent a range of moral cognitive features, including cross-cultural variability and the simulation of signature moral dilemmas.
WHEN: MONDAY 21ST OCTOBER
WHERE: F23 LEVEL 5 FUNCTION ROOM
Copyright © *|2019|* *|School of HPS*, All rights reserved.
Our mailing address is:
*|hps.admin at sydney.edu.au|*<mailto:*|hps.admin at sydney.edu.au|*>
Want to change how you receive these emails?
You can update your preferences<*|UPDATE_PROFILE|*> or unsubscribe from this list<*|UNSUB|*>.
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the SydPhil