IDA Model Overview

1. IDA Model Overview Stan Franklin and the Conscious Software Research Group

2. Stan Franklin Art Graesser* Chip Ordman Sri Satish Ambati* Ashraf Anwar* Ramesh Aitipamula* Igor Beliaev Myles Bogner* Deirdre Bolden Sidney D’Mello Scott Dodson* Arpad Kelemen* Ravikumar Kondadadi* Sarveshwar Reddy Kuncha Lee McCauley Irina Makkaveeva* Gurumoorthy Nagasubramanian* Aregahegn Negatu Brent Olde* Vani Priya Uma Ramamurthy Hongjun Song* Alexei Stoliartchouk Matthew Ventura Yun Wan* Zhouhua Zhang* former member off to better things* The Conscious Software Research Group

3. Autonomous Agent An autonomous agent is a system • situated within and a part of an environment • that senses that environment • and acts on it, • over time, • in pursuit of its own agenda • so as to effect what it senses in the future.

4. Examples

5. I A D Internet IDA: an Intelligent Distribution Agent Dialogue with sailors Read personnel data Check job requisition lists Enforce Navy policies Choose options to offer members Negotiate with them about jobs Telephone Detailer

6. Automomous Agent GW Theory Conceptual Model Computational Model IDA: a “conscious” software agent IDA

7. Global Workspace Theory • A psychological theory of consciousness • The nervous system is a distributed parallel system with many different specialized processors • Global workspace contains a coalition of processors • Broadcasts globally to all other processors • Recruit other processors needed for any degree of novel or problematic situation • Explains limited capacity and seriality

8. Contexts at work

9. Modeling Cognition • Situated (embodied) cognition—Varela, Thompson & Roach • Long-term working memory—Ericsson &Kintsch • Perceptual symbol systems—Barsalou • Memory via actions—Glenberg • Global workspace theory—Baars • Cognitive architecture—Sloman

10. Why a ‘Conscious’ Agent? • Flesh out the theory with detailed architecture and mechanisms • Hypotheses for cognitive scientists and neuroscientists • Produce flexible, adaptive, human-like software • Want smart agents? Model them after humans.

11. Modules and Mechanisms • Perception—Copycat Architecture—Hofstadter • Action Selection—Behavior Net—Maes • Associative Memory—Sparse Distributed Memory—Kanerva • Episodic Memory—Case-based Memory • Emotions—Pandemonium Theory—Jackson • Metacognition—Fuzzy Classifier Systems—Holland, Zadeh • Learning—Copycat Architecture, Case-based Reasoning • Constraint Satisfaction—Linear Functional • Language Generation—Pandemonium Theory • Deliberation—Pandemonium Theory • “Consciousness” —Pandemonium Theory

12. IDA’s Architecture Metacognition Database Perception Constraint Satisfaction Deliberation Negotiation Problem Solving Behavior Net Expectation & Automization “Consciousness” Perception Associative Memory Episodic Memory Emotions

13. A continuing iteration of a cognitive cycle of activities involving: Perception Working memory Transient episodic memory Long-term associative memory Consciousness Action selection Motor activity Processing in IDA

14. PRECONSCIOOUS PERCEPTION PERCEPT TO PRECONSCIOUS BUFFER LOCAL ASSOCIATIONS COMPETITION FOR CONSCIOUSNESS CONSCIOUS BROADCAST RECRUITMENT OF RESOURCES SETTING GOAL CONTEXT HIERARCHY ACTION CHOSEN ACTION TAKEN Cognitive Cycle

15. Cognitive Cycle Processing • Hypothesis— Like IDA’s, human cognitive processing is via a continuing sequence of Cognitive Cycles • Duration— Each cognitive cycle takes roughly 200 ms with steps 1 through 5 occupying about 80 ms. • Overlapping—Several cycles may have parts running simultaneously in parallel • Seriality— Consciousness maintains serial order and the illusion of continuity • Start— Cycle may start with action selection instead of perception

16. Levels of abstraction • High level • behaviors • message type nodes • emotions • metacognitive actions • etc. • Low level • codelets

17. Codelets • Small pieces of code each performing a simple, specialized task • Acts as a demon, always watching for a chance to act • Most subserve some high level entity, e.g. • behavior • slipnet node • metacognitive action • Some codelets work on their own, e.g. • watching for incoming mail • checking for time and place conflicts • Codelets do almost all the work • IDA is a multi-agent system

18. Perception via a Slipnet Norfolk nor NRFK norfolk Norfolk . . . San Diego Jacksonville Miami location information request preference acceptance

19. Sparse Distribute Memory — Boolean Space — dim = N (enough to code features) bit vector Job List Outgoing Message Focus Sailor Data Working memory Perception Behavior Net Negotiation Deliberation Associative Memory

20. Coalitions and Consciousness • Coalition manager • Spotlight manager • Broadcast mechanism

21. Drive to Acknowledge A Behavior Stream Activation from drive Send an acknowledgement Compose an acknowledgment Find an email address Find and move a template From the Sidelines Activation from the environment, external or internal

22. Broadcast Behavior net templates Behavior Net in Action Behavior net Work Space Stands Side lines Playing field

23. Associative Memory Working memory Job List Outgoing Message Playing Field Stands “Consciousness” in Action Focus

24. Deliberation • Faced with a goal or problem • Imagine possible plans or solutions • Scenarios • Routes • Internal virtual reality—Dawkins • Evaluate them • Using reason • Using emotions • Choose among them

25. IDA’s Deliberation • Create scenes • May require objects, actors, concepts, relations, frames • Organized around events • Build scenarios as sequences of scenes • Choose between scenarios, discarding some Using Barsalou’s perceptual symbol systems as a guide

26. Associative Memory Working memory Detach Date Job List Leave Time Playing Field Detach Date Leave Time Stands Leave Time Detach Date Deliberation in Action Focus

27. Ideomotor Theory William James (circa 1890) ----- Bernard Baars (1988) Voluntary vs non-voluntary action Theory of voluntary action • Proposers—propose a course of action • Objectors—raise objections to such a course of action • Supporters—lend support to such a course of action • Auctioneer—wields the gavel

28. Ideomotor Theory in Action Idea pops to mind (proposer)—no objection (objector)—do it Objection (objector)—don’t do it Objection then support (supporter) do it Different proposal—no objection do it Different proposal—original proposal—no objection—do it Last unopposed proposal is acted upon

29. Web and Email Addresses • Stan Franklin • franklin@memphis.edu • www.cs.memphis.edu/~franklin • “Conscious” Software Research Group • www.cs.memphis.edu/~csrg • (Addition to Dr. Franklin’s slide --) • Edward Ordman, edward@ordman.net, //www.ordman.net • Prof. Lee McCauley, mccauley@memphis.edu