Tutorial T3: Empirical Principles for LSD Specification

Context: modeller is observer of a system over a period of time, possibly able to intervene in operation of the system, whose objective is to conceive and construe the behaviour (cf James's activity of the system).

Conceive: all behaviour is conceived - looking at certain things in a certain way, presuming some integrity, reliability and persistence in what is observed. Choosing what to observe and how to observe it.

Construe: identify agency and attribute activity to agents, interpret effects of action.

Can draw on past experience, legacy of other people's experience

No one correct way to do this: pragmatism. Can take the perverse view that are responsible for nothing, that all our actions are pre-ordained etc. Every agent could be a puppet of a super-agent. An agent could be present always, but only intermittently active (otherwise a silent onlooker)

Significant that analyses of this kind can provide some basis for predictions about a system, but fall far short of prediction of behaviour such as physical theories typically provide. Attribution of change vs. prediction of change.

Assumptions about the super-agent (the modeller) are also relevant: power to recognise and remember, to recognise that you are acting, and acting with intention.

Illustrative examples: distinguish between

can I brake?, and can I stop at a red light?

Useful exercise in this connection is to frame experiments that might be used to perform the identification of agents, and associated classification of observables. Some possible suggestions for such experimental procedures given overleaf.

1. have I any agency in a system?

? is there an observable that I can change

[vacuously true if we have enough imagination about the world]

cf. ? is there anything I can do about the situation

 2. is there an agent present? ? does any observable change in a fashion uncorrelated with my actions

[what is uncorrelated: clearly observer dependent, influenced by what models of dependency we have access to cf trigonometry]

 3. am I the sole agent present? ? is every change to observables correlated with changes that I initiate 4. is a particular agent present? ? is there a correlation between the presence of certain observables and disposition to change in the system

[this type of experiment establishes observables characteristic of an agent identity]

 5. does a particular agent exercise conditional control over a particular observable? ? is there a correlation between the presence of observables characteristic of an agent identity and disposition to change for particular observable

[this type of experiment establishes handles associated with an agent identity]

 6. does a particular observable influence the behaviour of a particular agent? ? is there a correlation between the presence of observables characteristic of an agent identity and disposition to change for particular observable

[this type of experiment establishes oracles associated with an agent identity: it could perhaps be construed as a dual to 5 if we consider exercise conditional control over and influence the behaviour as synonymous.]

 7. is a particular observable constant or subject to variation? ? is the value of observable ever seen to change

[this type of experiment distinguishes between kinds of system parameters]

 8. how many agents are present? Many experimental strategies to offer insight:
  a. particular changes to observables are characteristic of particular agents

[door is unlocked - who has a key?]

b. simultaneous change to several observables that not correlated indivisibly

[tap on window, light is switched on, and a coal rolls out of the fire]

c. pattern of system behaviour characteristic of particular combinations of agents

 [2-body vs. 3-body system / discovery of Pluto]