Friday 25 January 2013

Dostoevsky on deciding what to do

There is an intriguing view of human decision-making and freedom in Dostoevsky's Notes from Underground, part 1, chapters 7 and 8. Here, I shall analyse an argument from chapter 7.

Dostoevsky starts with a question. Why do we, again and again, knowingly and deliberately act in ways that are contrary to our own advantage? He promptly moves on to another question. What is a person's advantage? Then he claims that there is an advantage that cannot possibly be included in any catalogue of advantages.

The advantage that cannot be included in any catalogue is that of making an unfettered choice, in defiance of any careful computation of advantage. The argument may be reconstructed as follows. (This reconstruction goes beyond what Dostoevsky says, in order to secure the argument against some obvious logical objections.)

1. Suppose that a person, D, is aware of the contents of a supposedly complete set, S, of advantages to him or her, along with information about their relative importance, about how to obtain the advantages, and about the possibilities for obtaining various combinations of them. "Advantages to D" has a broad meaning. It may include advantages of benefiting others, at no obvious gain to D. There is no suggestion that D need be selfish in making the best possible selection of advantages from S.

2. D can now make a careful computation of what to do, in order to maximize the net advantage to D.

3. D can also exercise freedom, by going against the result of the computation.

4. The exercise of freedom is in itself an advantage to D, but it cannot be the result of the computation, otherwise it would not be an assertion of D's freedom. D could freely decide to act in accordance with the result of the computation, but that would not be the kind of unfettered freedom that is required here. It would not assert D's ability to live unconstrained by such computations.

This argument does leave space for action in defiance of the result of the computation to be included in S. But defiance could not coherently appear as all or part of the result of the computation. Its inclusion in S would therefore be idle.

Suppose, first, that the computation produces a single recommendation, to act in defiance of the result of the computation. Compliance with the recommendation would amount to defiance of it, and defiance of it would amount to compliance with it. (There would be the additional, substantial, difficulty that D would not know what to do. "Act in defiance of a recommendation to eat healthily", would give D an idea of some specific action. "Do not do what this sentence tells you to do", would give no idea of any specific action.)

Now suppose that the computation produces several recommendations, say "Eat healthily", "Move to another city", and "Act in defiance of the result of the computation". If we read this list as a conjunction, as I think should, we find that D cannot comply with all conjuncts. Suppose that D eats healthily and moves to another city. Then if D complies with the final conjunct, it can only be by defying that conjunct, since that is the only remaining way to break the terms of the conjunction. If, on the other hand, D defies the final conjunct, then D must comply with the whole conjunction, which must mean complying with the final conjunct, as well as with the other two.

The one apparently coherent option would be to defy either or both of the first two conjuncts, and thereby comply with the third. But on closer inspection, we can see that this would not work either. The reason is that it would be known in advance that "Act in defiance of the result of this computation" would amount to "Discard at least one of the specific prescriptions". Given that acting in defiance of the computation was considered to be an advantage, a member of S, this discard would be required in order to yield the optimal solution, if the prescription to act in defiance of the result were to be part of the result. Therefore, the discard of at least one of the specific prescriptions would form part of the calculation, before the result was given. (The prescriptions to be discarded might be chosen by some rule, or at random.) But then the result would be a conjunction of the remaining specific prescriptions and the prescription to act in defiance of the result. Again, at least one of the specific prescriptions would have to be discarded, in order to achieve the optimal result while still keeping the prescription to defy as part of that result, and this too would have to form part of the calculation. We would continue until only the prescription to act in defiance of the result was left. But as already noted, that would lead to incoherence.

We may therefore conclude that while the prescription to act in defiance of the result could be included in S, contrary to what Dostoevsky asserted, it could not coherently feature in a prescription of what to do in order to maximize advantage, so its inclusion in S would be idle.


  1. I think Dostoevsky sees this model of decision-making - the careful computation of advantage - as being inherently limited to choices that the rational person would accept as beneficial, "the obligation to want only what is sensible". The gratification of the narrator's masochistic desires could not therefore appear in the set of advantages because Dostoevsky's rational person would not see degradation and humiliation as benefits.

  2. Thank you Isabel, that is interesting. If we look at it your way, and incorporate the distinction between what rational people would want and irrational desires, before considering what a particular individual (D, in my analysis) actually wants, that puts the whole thing in a different light. I would have to re-work the analysis, perhaps along these lines.

    There is a class, F, of advantages, broadly, the fulfilment of rational desires. Then there is a class, G, of disadvantages, broadly, the fulfilment of irrational desires. Any member of F, but no member of G, can appear in the recommendation that results from a rational computation. (Incorporating "any member of F" at this stage rules out the inclusion in F of something that cannot appear in the result, the option that I contemplated in my original post, as idle inclusion.)

    Then the fulfilment of a desire to act in defiance of the result of a rational computation cannot appear in F, because that desire could not be a rational desire. It could not be a rational desire, because the result of a rational computation is to go as far as possible in fulfilling rational desires, and all of the included fulfilments would be drawn from F. The fulfilments would not include any that were drawn from G, because it is irrational to pursue irrational goals. (If the result of the computation did include anything drawn from G, the agent could defy the result of the computation by not acting to achieve some fulfilment that was drawn from G. And that would be a rational way to act, because it is rational not to pursue irrational goals.)

    The fulfilment of a desire to act in defiance of the result of a rational computation must therefore appear in G, if it appears anywhere.

    So your distinction between rational and irrational goals makes the analysis more straightforward. There is no need to contemplate iterations that gradually eliminate all desires apart from the incoherent one.