REMEMBERING WHO WE ARE - HUMAN MEMORY FOR ARTIFICIAL AGENTS
A one day symposium on the 29th March 2010
In conjunction with the AISB 2010 Convention
De Montfort University, Leicester
(The symposium is supported by the European FP7 Project LIREC)
Memory gives us identity, shapes our personality and drives our reactions to different situations in life. We actively create expectations, track the fulfilment of these expectations and dynamically modify our memory when new experiences demand it. Yet up to date, many important social aspects of human memory (for instance, emotional memory and episodic memory) to artificial intelligent agents have not been given much attention. The challenge might lie in the amount of memories one can have in a life time. Take a narrative agent for example, how can we generate a lifetime’s worth of memories for this agent? Can we easily record human experiences for this purpose? What trust and privacy issues will this entail? On the other hand, without this type of memory, can the agent generate believable life stories given that it is what colors our lives in retrospect? For an agent that continuously interacts with users or other agents, how can we design it with the capability to generate memories worth remembering in its lifetime? How can the agent record experiences of others during interaction? Can the agent maintain its relationship with others without any information about its past experiences with them? Artificial agent researchers have been constantly coming up with computational cognitive models inspired by the human brain to create characters that are more natural, believable and behave in human plausible ways. However, memory components in these models are usually oversimplified. Memory components which have been widely accepted and modelled are the long-term memory including procedural and declarative memories, the short-term memory and the sensory memory. What about the more ‘socially-aware’ memory which allows us to be effectively involved in social interactions and which fundamentally supports the creation of our life stories including the significance of events and their emotional impact? It is important to review artificial agents without this kind of memory particularly those designed for social interactions, and reflect on the effects of this shortcoming. Additionally, many existing models do not take into consideration the bio-mechanisms of human memory operations such as those involved in retrieval and forgetting processes. The most commonly adopted approach to forgetting is decay but the human brain performs other processes such as generalisation, reconstruction and repression to list a few. This symposium offers an opportunity for interdisciplinary discussions on human-like memory for artificial agents including organisational structures and mechanisms. We hope to bring together memory researchers, psychologists, computer scientists and neurologists to discuss issues on memory modelling, memory data collection and application to achieve a better understanding of which, when and how human-like memory can contribute to artificial agents modelling.
Topics of interest include but are not limited to:
Submission We are seeking submissions of original papers (up to 8 pages) that fit well with the symposium theme and topics. Papers should be submitted through the EasyChair system. You will have to register with EasyChair if you do not already have an account. Please submit your paper in PDF format (according to the AISB 2010 formatting guidelines - templates available on the AISB 2010 convention website). All submissions will be peer reviewed. Authors of accepted contributions will be asked to prepare the final versions (up to 8 pages) for inclusion in the symposium proceedings. At least one author of each accepted paper will be required to register and attend the symposium to present their work.
Cyril
Brom, Charles University Prague
http://www.macs.hw.ac.uk/~myl/AISBRWWA.html
Dr. Mei Yii Lim Dr. Wan Ching Ho |