Earlier, however, van Gelder made a different challenge to computationalism [15]. He argued that the conceptual anchor for understanding cognition shouldn't be the Turing machine, but rather the Watt governor, a simple device that, through its dynamical properties, regulates the speed of a steam engine. The Watt governor performs its function without any use of algorithms, representations, etc.; any such paraphernalia would merely impede its elegant means of co-ordination. Perhaps, then, we should also see minds as systems in coupled correspondence with their environments, rather than as discrete, representational, symbol manipulating systems.
This is an example of an objection to computationalism that is not handled well by taking the transparent approach. That is, it seems very unlikely that systems such as the Watt governor will be best understood in computational terms. If computation is essentially intentional (about something), and if Smith is right in claiming that intentionality requires both connection with and disconnection from one's subject matter [14], then the close coupling between the Watt governor and its engine which van Gelder extols is exactly the reason for relegating it as less than intentional and thus less than computational. Simple connection and simple disconnection come for free in the physical world; negotiating between the two to maintain some abstract correspondence is a more sophisticated, and (at least sometimes) computational, achievement.
Since it seems unlikely that our notions of computation can be expanded to include the Watt governor, a different kind of reply is needed. However, one can use the way in which transparency failed to work as an indication of the basis of a proper response: if the Watt governor is too simple to be computational, perhaps it is too simple to be a good conceptual anchor for cognition [3]. However, there is a slightly different Watt governor, or a description of the Watt governor that is different from the usual characterisation, which might have the complexity necessary for being both a conceptual anchor for cognition and for warranting an extension of the notion of computation. Consider a Watt governor that can become temporarily disengaged from the engine it is regulating (such a governor might be useful in the same way that a clutch is useful in a car). If the Watt governor becomes momentarily disengaged, it doesn't instantly stop spinning. If it did, it would adopt a configuration which it should only be in when the engine speed needs a maximal increase, which would be inappropriate in the case we are considering. The Watt governor needs some way of distinguishing temporary disengagement from the (relatively rare) case of the engine needing maximal acceleration. This is provided naturally in the dynamics of the governor: even when it is disengaged, its inertia and momentum mean that it will only slowly, and continuously, reduce its speed and thus the angle of its arms. Thus, when re-engagement occurs, it is likely that there won't be an enormous mismatch between the speed of the governor and the speed it should have, given the engine speed at the time of re-engagement. The suggestion here is that the inertia of the governor maintains a correspondence between the speeds of the engine and the governor even when they are disconnected. Is this a kind of negotiation between phases of connection with and disconnection from the subject matter, which Smith claims is at the heart of intentionality? 4 If so, or if one can imagine modifications to the governor that allow notions of intentionality to apply without destroying the Watt governor's purely analogue, dynamical nature 5, then one would have grounds for extending the notion of computation and taking the transparent computationalist line as a way of responding to van Gelder.