Originally published in Ethics 84,4 July 1974 as "More on Rewards and Reinforcers: a reply to Michael Schleifer"


ARE REWARDS REINFORCERS?

On Probabilizing Behavior
©1999 Edward G. Rozycki
RETURN

edited 12/07/03

If Professor Schleifer's reading of my article 1 is the common one, then, inasmuch as clarity of intent is a measure of quality, I must own up to having done an execrable job. This is not, however, to make any philosophically important concessions, for, in fact, what Schleifer gives as my, argument is not my argument. Nor is my argument - which is ignored entirely by him - premised on any considerations vis-a-vis ordinary language notions of reward and punishment. Schleifer's criticism misses the mark by a wide margin.

My argument -- which I will presently review -- was that there is a logical incompatibility between the language of voluntary behavior and that of reinforcement theory. This conclusion is based on a consideration of the mathematical foundations of operant theory; to wit: the partitioning of behavior types needed to satisfy the definitional requirements of the probability calculus destroys the complexity of description needed to distinguish voluntary from involuntary behavior. Since conditionable behavior must be probabilizable, voluntary behavior cannot be conditioned. (I do not -- as Schleifer insists -- arrogate the term 'act' to humans and restrict 'behavior' to what animals do. If that is the impression one gets, I must again plead faulty exposition.)

If my argument is correct, it does not follow -- nor did I argue -- that the results of research in the control of autonomic functions has no relevance to our moral and educational concerns. What does follow is that this relevance cannot be pleaded on considerations derived from the language of reinforcement. I did not make reference to the research Schleifer cited in his original article 2 for two reasons: I was not concerned so much with that research as with his attempt to argue from notions of reinforcement to those of voluntary behavior; where the relevance of certain research may be under question, to allude to such research is to presume that relevance.

I will attempt to -- briefly -- three things: clarify the dispute vis-a-vis the notions of reinforcer and reward; restate the argument that I can justly be held accountable for; and, finally, perform the ungracious but necessary chore of correcting some of Schleifer's more important misconstructions.

Rewards And Reinforcers

I suppose that recourse to "ordinary language" -- although a traditional move in philosophical argumentation -- is somewhat disingenuous. I doubt that any of us is much concerned with what just anybody would say about reward and punishment; but rather, in practice, we restrict ourselves to what knowledgeable persons, judging carefully and adept at a perhaps finer than everyday English, would say. This is not to say that the appeal to ordinary usage is fraudulent. It might be more precisely put as an appeal to the usage of those disinterested in a particular point under dispute. But if the point is general enough, there may be no room for disinterestedness. Partisanship in such a case would define different communities of usage.

In footnote 163 I stated a belief which - although not bearing on the immediate issue - serves to orient my approach to disputes such as those in which Schleifer and I are engaged. It is that the recourse of the philosopher to ordinary language is not as to authority but as to an intersubjectively reliable category system. (Despite this footnote, it still seems to Schleifer that I take ordinary language as sacrosanct.) We normally presume community of usage and the concomitant reliability of the language; a dispute often indicates a lack of such.

The question can be put simply: given the use of the same term by different communities of interest, for example, philosophers and psychologists, are they employing the same concept? This is a common -- perhaps misleading -- way of asking whether psychologists adduce the same considerations as do philosophers in determining whether or not the term applies in a particular situation. If they do, then the concepts are the same. If they are not the same, they may be logically related in that all the considerations adduced by one group are included among those of the other. There may be a purely contingent relationship, where the same term is applied by different communities of interest vis-a-vis sets of considerations that are logically compatible.In this case, we may find that we can speak of having the same thing treated relative to different aspects; for example, reward treated in its psychological as contrasted with its ethical aspect.

There is a fourth circumstance which I think obtains with respect to those concerned with moral problems and those working with conditioning phenomena: namely, that, despite common terms, the concepts are incompatible, that is, a condition necessary to one conception obtains only if one necessary to the other does not. (I think that it is often not much more than respect for the academic degree that leads us to say of someone that he has a different concept rather than than he is misusing the language.)

A major element of the concept of reward that Schleifer alludes to is that a reward must be something which produces pleasure of some sort.4 Similarly, punishments seem to need to have displeasing effects.If something like the formula, "X is a reward only if someone or something finds X rewarding," catches Schleifer's notion of reward, then by way of contrast, the concept of reward that I alluded to could be partially caught by, "X is a reward only if X is given in reward." The reason for the present dispute should thus be very obvious: for Schleifer, effect is central; for me (and, I believe, many others) intent and uptake are. The question is, which concept is more suited to the expression of moral (and legal and educational) concerns?

Neutral Transfers

Let us call the following situation a neutral transfer: at time t1 Harry has an envelope of money in his hand; at time t2 John has that same envelope of money in his hand. Let us also bring into consideration some behavior of John's -- call it "M." The neutral transfer was reinforcing for M under the circumstances only if the probability of M has increased (assuming -- for the moment -- that it is coherent to speak of such). But was the neutral transfer a reward? What considerations might be adduced in determining the status of a particular neutral transfer vis-a-vis its being a reward?

We must know whether Harry gave it to John for M-ing, otherwise it might be a gift or a loan (among other things);(b) whether Harry intends that John have it, otherwise it may be a theft (among other things); 5 (c) whether John may dispose of the money as he sees fit, otherwise it may be a consignment; or (d) whether Harry need want John to have it, otherwise it may be merely a payment, which Harry is obliged by contract to render. Gifts, loans, thefts, consignments, and payments are not rewards. What follows so far is already agreed to by Schleifer; a reinforcer need not be a reward.

On the other hand, to know that the money is a reward, it is sufficient to know that Harry wants John to have it and gave it to him because he M-ed; that Harry's giving it was, so to speak,an act of grace and not an obligation -- this distinguishes it from a payment or an extortion; and finally, if the concept is to have any moral import at all, that John is capable of recognizing Harry's intent and appreciating the circumstances as determining it to be a reward rather than something else. 6 No effect on John's behavior is necessary to identify it as a reward; nor need John find it rewarding (pleasurable) to receive the money.The same reward may provide more or less or no incentive, depending on the circumstances of the recipient.

Although I risk Schleifer's censure for linguistic dogmatism, I am presenting what I take to be facts about the concept of reward, at least as used by those for whom the distinctions between reward, incentive, payment, compensation, remuneration, prize, requital, tribute, reparation, redress, etc., are important. Because of such considerations,I would insist that rewards are institutions -- not necessarily human institutions,for other species may be capable of recognizing intent, appreciating the existence or absence of obligations, and so on. But why is it that we say, for example, that we reward our pets for good behavior? Why do we not say that we pay them? Why is not my dog's dinner part of a salary given him for providing entertainment to my children?

I think it is clear that in saying that we are rewarding animals, the term 'reward' does not admit of such contrasts, and only to the extent that we would think of animals as persons, that is, as possessors of rights and obligations, would we find such contrasts viable. To say that we are rewarding animals is to extend that term beyond the limits of moral significance. That we do this with 'reward' rather than with 'payment' derives -- I speculate -- from our relationship to our pets or animal charges as that of a sort of benevolent despot: "X rewards Y with Z (for M) " admits of no obligation for X to give Y to Z; and it is by the grace of X that Y receives Z.

If we allow that rewards on Schleifer's conception are all reinforcers, it follows from the preceding discussion that not all rewards on the conception that I have sketched out are reinforcers, for the concepts are different, although the term is the same. This conceptual difference disallows any "translation" from the rewards of reinforcement theory to the rewards of ethical concern.7 However, from the preceding discussion it does not follow that the two concepts are incompatible; there is yet considerable room for contingencies to be discovered between the two concepts. This can be seen as follows: let Schleifer's concept of reward define the set RS of all instances of reward identified by adducing the considerations relevant to that concept; similarly define RR relative to the concept I have attempted to explicate. If we disregard the set RS-RR, morally irrelevant rewards, and also the set RR-RS, non-reinforcing rewards, we still have the intersect of RR and RS, available for reinforcement research. It will take a different kind of argument to show that RR intersect RS is empty, that is, that there are no morally relevant reinforcers.

Probabilities And Behavior

Schleifer takes umbrage at what he mistakes again as my claim that the psychologist's use of the terms 'reward' and 'punishment' in describing animal research is degenerate usage. That statement was that the psychologist using a BP (behavior partition) -- not just any psychologist -- was of necessity (by antecedent argument) using a degenerate form of English, 8 in that a set of behavior types that constituted a partition could not allow the distinction between voluntary and involuntary behavior. (Nor did I restrict this claim to those engaging in animal research.) My argument that voluntary-act types are not partitionable is the core of the essay.

The considerations that lead up to it are these: something is a reinforcer only if it changes the probability of some behavior in some circumstance.The coherence of talk about the probabilities of behavior depends upon the theoretical possibility of sorting out behavior into mutually exclusive types, 9 lest closure for set-theoretic operations necessary to the definition of probability-distribution functions (and, I might add, any measure function whatsoever -- for example, a tally) not obtain. 10

To say that some behavior is possible -- with this approach -- is to say that that behavior type belongs to or is a covering term for behavior types that belong to the set of behavior types upon which a probability distribution is to be defined. The criteria v1 and v2 were not -- as Schleifer would have it -- given to underscore the problem of the relationship of the actor's beliefs to the identity of his acts - that is a secondary consideration; rather, these criteria serve to indicate what act types must be included in the set of behavior types in order that some behavior be identifiable as voluntary. The argument is then that only sets of behavior types containing some 'trying to . . .' types can be adequate to identify voluntary behavior. But such sets are never partitionable. Therefore voluntary behavior is not probabilizable; consequently, not conditionable.

The simplest kind of BP is a dichotomous one containing, for example, 'X' and 'not X'. This is the kind we resort to informally, as when we count how many times someone jumped, or laughed, or gave a correct answer. We can imagine a series of experiments, E1, E2 ... for each of which a dichotomous BP is defined; for example, BP1, BP2, ... From the possibility of so doing it does not follow that the union of the BPs is a BP, just as we may tally separately but not together all the apples, fruits, shoes, and useful objects in a box. From my argument it follows that there can be no general reinforcement theory that applies to voluntary behavior. The immediately preceding considerations show that that argument is not countered by the possibility of informal tallies or restricted studies.

Final Remarks

No further purpose will be served by my continuing to identify and correct other misconstructions. I am satisfied that the nature of the problem has been clarified and my argument restated, if only in outline. Of course, I concede Schleifer all those points upon which we are really in agreement -- they are rather numerous. I recognize my responsibility in contributing to the confusion by having presented originally what I suspect is a rather oblique argument.

As regards that argument - which I think can be construed as a variation of the Free Will-Determinism problem - a speculative idea can be entertained: to the extent that we can control behavior, are we deprived of a good reason for doing so?



REFERENCES

1. Edward G. Rozycki, "Rewards, Reinforcers and Voluntary Behavior,"Ethics 84, no. I (October 1973):38-47.

2. Michael Schleifer, "Instrumental Conditioning and the Concept of the Voluntary," Ethics 82, no. 2 January1972):163-70.

3. Rozycki, p. 46.

4. He writes in his article, "Controlling Autonomic Processes," that all creatures capable of pain and pleasure can be rewarded and punished. I find it a stretch of the imagination to say I could punish an earthworm, although I could clearly administer adversive stimuli to it, but this is just another indication of conceptual incongruence.

5. To those more gifted than I at ordinary language analysis I leave the exposition of a really stunning array of contrasts.

6. It has been my experience that such appreciation has considerable effect on motivation.

7. Such a translation provides a crucial premise in B. F. Skinner's Beyond Freedom and Dignity (New York: Bantam/Vintage,1972). On page 31 he writes, "What the layman calls a reward is a positive reinforcer.' "

8. Rozycki, p. 45.

9. Understanding that we are not talking about simultaneity,the reader can appreciate what havoc is wrought with the language if behavior is described only by types which can satisfy the formula that if someone is X-ing, he is therefore nor Y-ing.

10. See, for example, early sections in Sterling K. Berberian, Measure and Integration (New York: Macmillan Co., 1965).

TO TOP