Distributed Cognition and Affect:Implications for AI :Distributed Cognition and Affect:Implications for AI 1st year AI+CS module


Introduction :AI+CS module 2 Introduction Cognition — impersonal mental behaviour (covers thought, perception, creativity, the possibility of personality, communication, action, planning — and is not restricted to humans). Isolated cognizers — in interaction with others and the environment; cognition is “in the head” and individuals are cognizers: conventional view. Distributed cognition — focus on the activity; do not presume “in the head” is usefully distinct from “out there” (sharing not fragmenting). Caution: fresh insight or mere re-description?


Additional characterizations :AI+CS module 3 Additional characterizations Distributed cognition arises because behaviour does not produce artefacts or traces which are “dead”. In cognitive terms such products are merely dormant. Perception is not concerned with ‘accurate’ representation (cf. a camera) but rather with interpretations aligned with contexts and situations. Objects and artefacts have affordances — recognizable functionalities recovered through perception (which is under active control, not simply ‘recording’). Thinking, etc., projects out into the environment, for the self and others to witness, comprehend, use (not just in communication).


Distributed Cognition :AI+CS module 4 Distributed Cognition Cognition is distributed over: Space Objects Others – human and non-human Time – historical and personal Affect is similarly distributed. (This will be covered later.)


DC over space :AI+CS module 5 DC over space Territoriality is an obvious example, which in humans we sometimes call ‘personalization’. Institutional spaces with special meanings. Memory distributed over space (food caches, PostIts and etc). Virtual spaces in computers can serve the same functions. The ‘Senator’ method of memory for speeches is in fact an internalization of an external distribution.


DC over objects :AI+CS module 6 DC over objects Tools are artefacts produced with DC projected onto them by the designer(s). DC is readable by the user and by others in terms of affordances and even cognitive processes (note conventionality — tools are not self-explanatory). Objects can be exploited for a function which is thereby afforded to another being who witnesses that utilization.


DC over others :AI+CS module 7 DC over others Teamwork and task sharing — control room scenario: Shared conception of tasks Observable participation which is readable in relation to shared tasks and goals Participant interchangeability Theory of mind. Teams involving non-humans. Reflexivity not essential (cf. cooperative behaviour in bees and other social insects). Distributed Cognition looks like communication.


DC over time :AI+CS module 8 DC over time Evolution of tools and mechanisms can be readable in the artefacts. Development of cross-head screw fixings is an example (cf. slot-head older forms). Cognitive processes can be embedded in artefacts, and over time this can become more specialised and effective (cf. navigational aids like sextants, astrolabes, etc., and calculators). Supposed evolution/development of computer application functionality (e.g. MSWord, etc.). Skill acquisition with a personal record.


Distribution of Affect :AI+CS module 9 Distribution of Affect Works the same way as DC: DA over space:— special places as emotional triggers &/or repositories DA over objects:— mementoes, photos, socially significant artefacts DA over others:— family bondings, shared experience in crowds and in teams DA over time:— religious custom, anniversaries Distributed Affect looks like communication.


Implications for AI :AI+CS module 10 Implications for AI AI needs to build models of Distributed Cognition and Affect which are situated — not build models of cognizers and try to put them in situations. Agent-based work looks most promising. Agents will need to have good models of humans and this raises ethical concerns (e.g. manipulation of affect by a robot toy). Notions of communication as situated activity make more sense in a distributed model — message model implies isolated cognizers.


The brain — universal? :AI+CS module 11 The brain — universal? Functional specification of the brain? Heart is for…. pumping blood Lungs are for…. blood/gas exchange Brain is for… ??? Physical and behavioural homeostasis General Cognitive Principles and the Sequential Imperative


General Cognitive Principles :AI+CS module 12 General Cognitive Principles GCP1: Sequentiality in behaviour is forced physiologically. Cor1: Sequence penetrates the corporeal boundary. Cor2: Sequence is semiotically free. GCP2: Cognitive entities are: Inherently atemporal Dual in nature GCP3: Behaviour is sequencing; Perception is de-sequencing.


General Cognitive Principles :AI+CS module 13 General Cognitive Principles GCP4: Learning serves the sequential imperative. GCP5: Attention is the management of the processes of sequencing and de-sequencing. GCP6: Affect is an attentional mechanism. GCP7: Cognition and affect can be distributed: in environment - space, objects, others (not just con-specifics) in time - historical, personal GCP8: Thought is the production of cognitive entities.


Introduction to SETI :AI+CS module 14 Introduction to SETI Assume it is worth searching — not because assessments tell us so; belief is enough. The question is not ‘Are we alone?’or ‘Where are they?’ but ‘How do we search?’ We need a strategic stance to guide: evaluation of existing SETI work, proposals for new SETI work. Targeted search proposal as illustration. Focus on universals, presuppositions, and conceptual coherence.


Strategic Stance :AI+CS module 15 Strategic Stance All sky survey versus targeted search. Theoretical presuppositions/coherence: Physics, engineering, biology, psychology, semiotics, … Magic frequencies/characteristics: Optical, ‘waterhole’, 56.8GHz, … Pulse, CW, FM, … Themes: Discovery Communications Protocol Message Design Social Implications Answers to Fermi’s question Preparedness to press SEND


Strategic Stance :AI+CS module 16 Strategic Stance All sky survey versus targeted search. Theoretical presuppositions/coherence: Physics, engineering, biology, psychology, semiotics, … Magic frequencies/characteristics: Optical, ‘waterhole’, 56.8GHz, … Pulse, CW, FM, … Themes: Discovery Communications Protocol Message Design Social Implications Answers to Fermi’s question Preparedness to press SEND


ETI’s Motivation :AI+CS module 17 ETI’s Motivation ETI intends to be discovered. Assume accidental ‘transmission’ is not relevant ETI adopts a strategic stance. ETI maximises discoverability: Artefactuality Value of reciprocity Communicative intent ETI targets its transmissions. TI needs to adopt an equivalent stance. TI targets search for targeted signals.


An Option — Pulsars as Beacons :AI+CS module 18 An Option — Pulsars as Beacons Pulsars as beacons provide: Directional alignment (targeting) Signal property (pulse rate) Artefactuality Details of ETI’s planet’s orbit etc. Edmondson, W.H. & Stevens, I.R. The utilization of pulsars as SETI Beacons. International Journal of Astrobiology 2(4): 231-271 (2003) Other schemes may be feasible.


Pulsars as Beacons :AI+CS module 19 Pulsars as Beacons Base case: Highly constrained but conceptually tidy ‘Perfect’ alignment: Pulsar —>> Earth > ETI ETI identifies our sun as prime habstar ETI transmits pulses towards our sun at a rate determined by the pulsar ‘behind our sun’ ETI uses lots of power and frequencies Communications protocol and message design not factors in discoverability Technologically ‘simple’ for TI


Pulsars as Beacons :AI+CS module 20 Pulsars as Beacons Extended case — I: Relax constraints (alignment precision; power/frequencies) Acceptable alignment within a 1º cone of space: Pulsar —>> Earth > < )ETI ETI identifies our sun as a candidate habstar ETI transmits pulses towards our sun at a rate determined by the pulsar ‘behind our sun’ ETI does not use lots of power and frequencies Communications protocol and message design not factors in discoverability Technology a significant factor for TI Calculations yield 792 alignments; 53 with ms pulsars (Turnbull and Tarter’s “Targets for SETI - I” list (2003) Astrophys.J. Suppl. 145, 181-198)


Pulsars as Beacons :AI+CS module 21 Pulsars as Beacons Extended case — II: Relax constraint on direction (Jean Schneider) Acceptable alignment within a 1º cone of space: Earth <<<<— < )ETI <<— Pulsar ETI identifies our sun as a candidate habstar ETI transmits pulses towards our sun at a rate determined by the pulsar ‘behind ETI’ ETI does not use lots of power and frequencies Communications protocol and message design not factors in discoverability Technology a significant factor for TI Calculations yield 1066 alignments; 60 with ms pulsars (Turnbull and Tarter’s “Targets for SETI - I” list (2003) Astrophys.J. Suppl. 145, 181-198)


Pulsars as Beacons :AI+CS module 22 Pulsars as Beacons ETI’s account: Adopt a strategic stance and work on issues of basic science and coherence of presuppositions Limited technical/social/energy commitment Target nearer plausible habstars Send some sort of message, but focus on: Discoverability Shared solution to shared problem Get on with lunch and life whilst waiting for response Be prepared to commit significant resources when response received


ETI’s ‘Pulsar pulses’ — 404? :AI+CS module 23 ETI’s ‘Pulsar pulses’ — 404? Problems with TI’s technology: RFI too disruptive Receiver technology inadequate Signal processing inappropriate [KLT] Signal processing faulty Dish characteristics limit search (Arecibo vs JB) General problems: ETI’s not there (this week; this century; ever) Magic frequency wrong — think about ETI’s radio astronomy [Eric Gerard], optical? Strategic stance wrong — ‘disproportionate interest’ in ILU unhelpful? Or should we be more thoroughly/assertively anthropocentric? [Paulo Musso]


Strategic Stance :AI+CS module 24 Strategic Stance All sky survey versus targeted search. Theoretical presuppositions/coherence: Physics, engineering, biology, psychology, semiotics, … Magic frequencies/characteristics: Optical, ‘waterhole’, 56.8GHz, … Pulse, CW, FM, … Themes: Discovery Communications Protocol Message Design Social Implications Answers to Fermi’s question Preparedness to press SEND


Presuppositions and coherence :AI+CS module 25 Presuppositions and coherence Physics Universal Engineering Universal? Biology Universal?? Psychology Universal? Semiotics Universal?


The brain — universal? :AI+CS module 26 The brain — universal? Functional specification of the brain? Heart is for…. pumping blood Lungs are for…. blood/gas exchange Brain is for… ??? Physical and behavioural homeostasis General Cognitive Principles and the Sequential Imperative


Semiotic universals :AI+CS module 27 Semiotic universals Arbitrariness Symbols Systems Conventionality Culture(s) Learning, change, errors Context Semiotic study Implications for message design


Conclusions :AI+CS module 28 Conclusions Sort out a coherent strategic stance. Targeted transmissions seem more probable and the search for them more worthwhile. Start looking using Edmondson & Stevens, and maybe optical as well as RF. Develop new targeting concepts. Implications for study of cognition/intelligence are profound: SETI means, in essence, looking for intelligences we can imagine on the basis of well-grounded assumptions. Anthropocentrism, or the search for ILU, is an antidote to SF because we can assume human intelligence is the non-unique manifestation of some general notion of intelligence.


Cognitive Science additional reading :AI+CS module 29 Cognitive Science additional reading “An Introduction to Language”, V. Rodman, R. Rodman, N. Hyams. Thomson. 2003. “The Honey Bee”, J.L. Gould and C.G. Gould. Scientific American Library. 1995 ISBN 0-7167-6010-X “Cognition in the Wild”, E. Hutchins. MIT Press. 1995. Maybe have a look at: “’Language’ and intelligence in monkeys and apes”, edited by S.T. Parker & K.R. Gibson. Cambridge University Press. 1990. “Where is everybody?” S. Webb. Copernicus Books. 2002. ISBN 0-387-95501-1 “Universe of the Mind”, Y.M. Lotman. Taurus. 2001. ISBN 1-850-430375-5