Simulating Attachment

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

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

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

Origins of Attachment Theory

John Bowlby, The Attachment TrilogyPsychoanalysis, Ethology, Evolutionary

Theory and CyberneticsEarly concentration on long separations,

loss, mistreatment and psychopathologyChanged hospital visiting practiceLater focus on Individual Differences

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

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

Maternal home behaviour prior to the SS

sensitivity-insensitivityacceptance-rejectionco-operation-interferenceaccessibility-ignoringemotional expressivenessrigidity(compulsiveness)-flexibility

The target behaviour

Attachment SS Subgroups vs prior maternal home behaviour

SS subgroupsMaternal Behaviour\ B1 B2/B3 C1 C2 A2 A1

Sensitivity 7.36 4.50 2.50 2.25 2.50 2.75

Acceptance 8.00 6.75 5.50 5.25 4.25 3.50

Co-operation 7.66 6.50 4.00 4.50 5.50 2.63

Accessibility 7.39 4.88 4.50 2.50 2.25 4.63

Lack of emotion* 2.17 3.88 4.25 2.75 6.50 6.00

Rigidity* 1.89 2.75 2.00 3.50 3.50 4.50

The target behaviour

Design methodology

Avoiding trivial solutions Whether to use data or theory to constrain the simulation Non-falsifiability

3 Problems:

Design methodology

Problem 1: Avoiding trivial solutions

The simulation is NOT trying reproduce superficial details of facial expression or body movement - like a Kismet robot might 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 BUT an abstraction of the target behaviour in a broad and shallow complete agent is TOO easy to reproduce

Design methodology

How to constrain the possible hypotheses space to exclude trivial solutions? Assume attachment styles are evolved, adaptive behaviours

Problem 1: Avoiding trivial solutions

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

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”

Design methodology

Problem 2: How to combine data, theory and AI techniques in the simulation - (Mook (1983) In defense of external invalidity)

Data and theories to be incorporated in the simulation

Data from the SS and other attachment studies

Bowlby’s theory

Distributed control architectures

Teleoreactive architectures

H-cogaff architecture

Theories of Executive Function - SAS

Machine learning algorithms (RL and ILP)

Design methodology

Problem 3: Non-falsifiability

Duhem, Auxiliary Assumptions

Popper, Falsifiability and Broad and Shallow architectures

Lakatos, Three kinds of falsification

Core assumptions and ad hoc assumptions

Progressive and Degenerative Problem Shifts

Design methodology

An unfinished simulation

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?

Architectural design issues

Bowlby’s theoryBehavioural 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

The H-cogaff architecture

Architectural design issues

The cogaff schema

Architectural design issues

Shallice and Burgess - SAS and contention scheduling

contention scheduling

SAS

Bowlby’s Behaviours represented in an infant-cogaff architecture

InternalWorkingModel?

Architectural design issues

Development of deliberative affordances or exploration and socialisation driven by deliberation?

Deliberationin re-union episodes

Development of partnership in planning as linguisticcompetencedevelops

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