Bernard Williams' doctrine that all reasons for a person doing something must be internal to that person- i.e. involve a motivation within himself rather than arise out of some Social Convention- appears easy to dispose off- at least on the assumption that human beings evolved by Natural Selection. This is because, assuming it is costly to have motivations which are subject to some calculus such that they inhibit or permit an action; it follows that for any given action, the Evolutionary Stable Strategy will militate for some proportion of actors not possessing an 'internal reason' but proceeding by mimicry, including lagged mimetic effects- i.e. for an external reason- to complete the same action. Indeed, Amotz Zahavi's work on warning calls and predator 'mobbing' shows how 'external reasons' can improve predator-prey outcomes and hence are trans-species Eusocial.
The same point would be true for artificial agents- e;g. those used in a simulation- provided it is costly to acquire information and apply reason to it and there is heterogeneity in terms of either information access or processing capacity amongst agents.
True, one could simply change the way one defines 'internal reasons' or 'motivations' such that we now speak of a desire to mimic or a desire to be lead or a desire to roll the dice and so forth. However, since these desires or motivations are encoded in Language, then- by Wittgenstein's argument against Private Languages- clearly these are 'external reasons' merely. The one doing the action may not even be aware of it, let alone have an internal reason for doing it. Certainly, it would also be a case of quod nescis quo modo fiat, non facis.
However, since Public Language has different granularity than either 'Mind stuff' or whatever Cognitive process determines action, and furthermore, since we know in advance that any given set of terms available to Public Language for use as explanans is either incomplete or misleading or both- it follows that no external reasons exist save by way of prejudice.
The literature on this subject, in so far as it relates to counter-factuals, misses the point that for agents who have evolved by Natural Selection, Uncertainty always exists as to which World we live in- this one or a counter-factual. Indeed, the Evolutionarily Stable Strategy- assuming it is costly to find out the answer- is to always remain in a state of 'buzzing blooming' confusion in this regard, at least for some proportion of the population. Philosophy has no Archimedian point here. On the contrary, opening its jaws to digest Game Theory it becomes merely the Ouroburos of the latter's shed skin.
No doubt, those familiar with the literature may think I'm missing the point. Perhaps, Williams is referring to ideal agents with infinite and instantaneous computing power? The problem here is such agents would have a type of Theory of Mind which would disintermediate Language and the essentially linguistic distinction between 'internal' and 'external'. Indeed, it is far from clear that the word 'reason' would retain any utility.
Alternatively, we may posit some special barrier to perfect Theory of Mind. But, in that case, Language would be entirely strategic because arising wholly from that barrier and the reason for it. Here again the distinction Williams makes becomes wholly hypocritical and empty of Philosophic content.
What about William's criticism of 'external reasons?'
The answer, of course, is that the agent believes he's hit on a 'cheap talk' ploy in a particular sort of game. To turn it into a 'costly signal' he may pretend to be undergoing some terrible inward struggle productive of the costive truth in question. Indeed, there's a Kavka's toxin type twist to this such that he has an incentive to convince himself that this is what is genuinely happening.
A bit like Bilgrami's Gandhi.