Lecture T1:

Concurrent Systems Modelling: Agentification, Artefacts, Animation
 

T1.1. A Hierarchy of Agent Models
 

commonsense concurrent systems

0-agent, 1-agent and multi-agent systems
 

external observer special in that can experiment
 

an agent is typically apprehended this way

... it can transport us to an unfamiliar context for observation
 

transport into a new context?
 

cf. previously undiscovered subtlety within the same observational regime

("That's the first time that anyone's played N-K4 in this variation")

vs. patent change of observational regime ("I think my opponent must be dead")
 

irrevocable, chaotic, change in the mode of operation of the system

lost integrity of the observational framework cf.

normal railway operation aftermath a serious accident
death of the external observer


Empirical classification of systems and models ...

 
don't know that resurrection is impossible, BUT ...

choice of classification pragmatic

e.g. assume reliability and speed of computer + fixed mode of interpreting its response

cf. "that's a pretty looking number" "Good grief! - Solaris is slow today"


Classification of Systems

0-agent system

no capacity to surprise

fixed correspondence between model and referent

cf. conventional formal model

only View 3 agents in 0-agent systems

1-agent system
in an n-agent system, where n>0,  there is a View 2 agent:
i.e. an entity capable of creating surprise cf. Newton's apple

many entities can be viewed as View 2 agents typically

e.g. critical components of a system under failure

For Empirical Modelling on conventional computer

best example of a 1-agent system = generalised spreadsheet

spreadsheet user = external observer is the only View 2 agent


Multi-agent = reactive systems

components reliable BUT is system reliable?

external observer in a 1-agent model acts as a superagent

simulate View 2 agency on behalf of other agents

c.f. Abstract Definitive Machine / user-driven eden

... in this fashion Empirical Modelling represents multi-agent systems


T1.2. Separating Agency from the External Observer
 
 

agent exercises power over our power

... so my agency not as external observer

"I can't be the external observer at my own funeral"

... cf. agency of the external observer
 

so have conceptual separation of agency from external observer
 

"conceptual separation" unavoidable:

 cannot literally see ourselves as others see us


artefacts represent our experience to ourselves

communicate our experience to other observers like ourselves
 

? how identify observers like ourselves

observe and interact in a common environment,
see congruence wrt agency and observation
Objective communication?
? private or shared conviction

James: "our minds meet in a world of objects which they share in common"
 

? same access to world of objects

NO: I might be colour-blind.


Artefacts as and vs. Language
 

interaction prior to formal language wrt knowledge and communication

natural language has words as artefacts

A sign on paper or a noise uttered <-> state of the world

situated ("phenomenological"?)  use of language

BUT formal language refers to View 3 agents


T1.3. Projection of Agency

Two perspectives on systems:

 
a. identify with the agents, and consider what the external observer sees
concurrent engineering
mutual surprise, mutual conflict, in view of an external observer
objective viewpoint on inconsistent, incoherent and asynchronous views


b. identify with the external observer, and consider what an agent sees

experimental scientist
represent other agency as if it were our own
explanation of a phenomenon
accounting for observed behaviour of agents via intelligible metaphors
An alternative interpretation:
 
a. know what the components do,
how can I determine what the system does?
how to realise a particular system behaviour?


b. know the system behaviour

how express in terms of component behaviour?
how design components and protocols to realise it?
Role of Empirical Modelling?

a. exploiting artefacts for interagent communication
b. representing what agent is presumed to "experience"

Explanation in commonsense concurrency is

"put ourselves in the position of another agent"
extrapolate from our own experience
project our stimulus-response context onto other agents
 
Most significant element in projection
passing from one consciousness to another
projection of human agency onto inanimate objects common


Empirical Modelling

constructs artefacts to represent
primary access to insight and knowledge
for all kinds of external agency


Development of artefacts liberated by computer: an extension of language?
 

T1.4. Elements in the Representation of Multi-Agent Systems

In Empirical Modelling for concurrent systems:

the process of agentification ("identification / attribution of agency"
the construction of artefacts
the animation of system behaviour


T1.4.1. Agentification

= identify the agents in a system

= find and classify the observables to which are construed to respond
 

Creative and experimental process: no guarantee of success

Can we identify reliable patterns of stimulus-response behaviour?
Can we represent them metaphorically using artefacts?
Identifying View 1, View 2 and View 3 agency
are steps towards successful representation, BUT
there may be no theoretical prediction of total behaviour

T1.4..2. Artefacts

"artefact" = physical / computer-based model + interaction

Artefacts imitate

experiences we have
experience we believe other human agents have
experience we can project onto inanimate agents to reflect their stimulus-response patterns


Roles for Artefacts

  • representing and acquiring personal conviction
  • in communication between human agents
  • representing agency cf. construals
    • artefacts <--> experiment & observation + skill acquisition

    computer as an instrument mediating between

    observable (conceptual entity) and perceptual (perceivable entity)
    T1.4..3. Animation

    system behaviour = consistent expectations about patterns of interaction

    violated when the system takes us by surprise
     

    3 views of system behaviour cf. 3 views of agency
     

    Duality between Agency and Behaviour

    only View 1 agents, behaviour ill-constrained

    only View 3 agents, the behaviour circumscribed
     

    3 Views of behaviour

    View 1 behaviour: Agent identification

    Observables partially identified and classified

    behaviour of the system explicable BUT

    via interactions not preconceived

    cf railway accidents

    prosaic interactions between agents + catastrophic interaction
    View 2 behaviour: Animation via superagent
    The observables and agents explain all possible behaviours

    privileges for action in certain contexts identified

    intelligent execution of privileges to simulate behaviour


    View 3 behaviour: Automation of behaviour

    behaviour of the system is circumscribed

    automatic execution of agent protocols.


    Classification of system behaviour is (of course) empirical in nature
     

    Empirical Modelling techniques to represent behaviours ...

    LSD specification

    identifying agents

    analysing and classifying the observables

    identifying privileges beneath pattern of interaction
     


    Abstract Definitive Machine for animation

    automatic animation is possible for View 3

    if View 1 and View 2 agents need intelligent execution