1. Simulating Attachment z Why simulate attachment? z Origins of Attachment theory z The target behaviours to be simulated z Design methodology and demo z Architectural design issues to be investigated

2. Why Simulate Attachment?  It provides a target for a design process - by building cognitive architectures to perform certain specific tasks we better understand architectures generally  Reproducing in simulation specific patterns of attachment behaviour acts as a ‘test-bed’ for developing architectural theories of human information processing

3. Why Simulate Attachment?  the developmental trajectory has normative stages which show representational change  initially only need to simulate limited cognitive resources  linked with evolutionary, physiological, anthropological, AI, cybernetic and cross-species data and theory  can abstract attachment behaviour

4. Origins of Attachment Theory  John Bowlby, The Attachment Trilogy  Psychoanalysis, Ethology, Evolutionary Theory and Cybernetics  Early concentration on long separations, loss, mistreatment and psychopathology  Changed hospital visiting practice  Later focus on Individual Differences

5. The target behaviour z The Strange Situation Experiment arose from comparing Ugandan and US infant attachment behaviours z Involves 3 separation/re-union stages z Each new stage increases the amount of anxiety they produce z Four patterns of response z A key pattern is link between home behaviour of mother and infant and infant behaviour on re-union in the SS

6. Infant reunion responses in the SS:  Secure (B type) behaviour  positive, greeting, being comforted  Avoidant (A type) behaviour  not seeking contact, avoiding gaze  Ambivalent (C type) behaviour  not comforted, overly passive, show anger  Disorganised (D type) Behaviour  totally disorganised and confused The target behaviour

7. Maternal home behaviour prior to the SS  sensitivity-insensitivity  acceptance-rejection  co-operation-interference  accessibility-ignoring  emotional expressiveness  rigidity(compulsiveness)-flexibility The target behaviour

8. Attachment SS Subgroups vs prior maternal home behaviour  The target behaviour

9. Design methodology z Avoiding trivial solutions z Whether to use data or theory to constrain the simulation z Non-falsifiability 3 Problems:

10. Design methodology Problem 1: Avoiding trivial solutions z The simulation is NOT trying reproduce superficial details of facial expression or body movement - like a Kismet robot might z It is trying to simulate the causal mechanisms behind the behaviour, at the level of goals and action plans within a complete agent in a multi-agent simulation z BUT an abstraction of the target behaviour in a broad and shallow complete agent is TOO easy to reproduce

11. Design methodology z How to constrain the possible hypotheses space to exclude trivial solutions? z Assume attachment styles are evolved, adaptive behaviours Problem 1: Avoiding trivial solutions

12.  Concentrating on Secure (B type), Avoidant (A type) and Ambivalent (C type) behaviours as potentially adaptive responses  Disorganised (D type) Behaviour is unlikely to be adaptive, but inclusion of this phenomena remains a possible future constraint Design methodology Problem 1: Avoiding trivial solutions

13. Design Methodology  Taking an evolutionary/adaptive approach the differences in infant security in the Baltimore and Uganda studies suggests the following questions:  Are Internal Working Models that are used in moments of attachment anxiety in part formed in episodes centred on non- anxious socialisation and exploration?  What information might infants gain from frequent episodes of exploration and social interaction that they use in infrequent episodes of attachment anxiety?  “If my carer won’t socially interact on my terms at all then I am less secure and I must use my own actions to gain security”  “If my carer sometimes socially interacts on my terms then I am less secure and need to concentrate my efforts in eliciting a response”

14. Design methodology Problem 2: How to combine data, theory and AI techniques in the simulation - (Mook (1983) In defense of external invalidity) z Data and theories to be incorporated in the simulation zData from the SS and other attachment studies zBowlby’s theory zDistributed control architectures zTeleoreactive architectures zH-cogaff architecture zTheories of Executive Function - SAS zMachine learning algorithms (RL and ILP)

15. Design methodology Problem 3: Non-falsifiability z Duhem, Auxiliary Assumptions z Popper, Falsifiability and Broad and Shallow architectures z Lakatos, zThree kinds of falsification zCore assumptions and ad hoc assumptions zProgressive and Degenerative Problem Shifts

16. Design methodology An unfinished simulation

17. Architectural design issues  how goals are chosen and represented?  when goals are chosen how are consequent behaviours chosen?  whether SS behaviour is deliberative or reactive?  how skill acquisition, chunking, parsing, perceptual affordances relate to attachment?  when and how new subsystems come on-line in attachment stage changes?

18. Architectural design issues Bowlby’s theory  Behavioural systems from ethology: attachment, exploration, fear and sociability  Stages defined by available control mechanisms: reflex (0-3), fixed action patterns (2-12), goal correction (9-36), goal corrected partnership (24- ), (age in months)  Coordination and control mechanisms: chaining, planning, ‘totes’, IWM’s and language

19. The H-cogaff architecture Architectural design issues

20. The cogaff schema Architectural design issues Shallice and Burgess - SAS and contention scheduling contention scheduling SAS

21. Bowlby’s Behaviours represented in an infant-cogaff architecture Internal Working Model?

22. Architectural design issues Development of deliberative affordances or exploration and socialisation driven by deliberation? Deliberation in re-union episodes Development of partnership in planning as linguistic competence develops

23. Architectural design issues  A distributed control system that adapts with Re-inforcement Learning at each node, has a non-central, non-symbolic representation, given by the genes, and undergoes no qualitative change in representation  A teleoreactive system that adapts using Inductive Logic Programming, has a simple central symbolic representation given by the genes that undergoes qualitative change in representation