Artificial Intelligence and Values
Values can refer to money, power and influence.
Influence is the most likely value related to the interface between human and machine - there must be coherent interaction. Imagine a stacked system where degrees of power increases interactive breadth and depth, and this kind of access is related to relevant exchanges of money. Power can refer to levels of authorization and access to system capabilities, but power can also refer to the resources of directing and the vice of threat.
Some think that there are no values operative in technological design and organizational systems but there never was a value free judgment. Even Weber who addressed value-free research in the social sciences specified that it referred to the exclusion of subjective interests, not to objective standards.
Anyway, positivism is largely eschewed today, even if few people understand the relation of facts to values or values to reality.
The values of AI must refer to the incentive value of money and the directional value of power, at least.
First of all, value free is a notion that has long been discarded - decisions are determined by interests which can be evaluated.
The question is whether the interests behind decisions are responsible, moral- ethical or legal. So assuming they are, the relation between values and facts reduces to something like this:
If A causes B and B is wrong or does not work, don’t do A.
If A causes B and B results is a worthy outcome, do A.
Here facts and values mirror each other; to reject this kind of pragmatism is the height of arrogance, ignorance and irrationality but it is still very common.
The values at issue today are money, power and influence. If
A value causes B fact, then do that value.
Keep in mind that these values are limited in their structuring effect: money incentivizes but too much money may corrupt. Vice is evil.
Power directs or threatens but inappropriate power claims are antagonisms and cause fighting. Murder all the way down to harassment is evil.
Influence means follow the law, follow the rules, but strict demands for conformity leads to the avoidance of normative or institutional behaviors. Pushing and nudging people around makes for avoidance, alienation, and anomie, just the reverse of the spirit of freedom and equality.
So values and facts must be seen in the light of each other. Designing AI systems does not mean foreseeing all of the effects of these structuring, coordinating systems but it does mean evaluating their effects and being open to change. Institutional practices, especially those determined by technological innovations are often very resistant to addressing their inappropriate effects.