Search This Blog

Friday, November 12, 2004

The Spandrel of Virtue: Radical Behaviorism and the Science of Optimism


Abstract

Radical behaviorism as defined recognizes only behavioral observations that are observable and replicable, and is coextensive with purely inductive psychological principles. Recently, the incorporation of micro-behavioral or neural events into a radical behaviorism has permitted the construction of new theoretical principles for incentive motivation that integrate Pavlovian salience motivation with the associationalist principles of Skinnerian operant conditioning. This for the first time permits the operationalization of key inferred processes in psychology such as intrinsic motivation, hope, and virtue that had heretofore been resistant to empirical analysis. Moral or humanistic principles thus may be derived without loss from a behavioristic analysis considered alone.

Radical Behaviorism in a New Century

A common view in psychology is that a behaviorism is about 'surface' things, overt behaviors, contingencies, and reinforcers, and that behavior is isomorphic to patterns of environmental events perceived now and historically (through memory). But what counts as behavior is dependent upon the resolving power of the tools you use to observe it, much as the behavior of the cosmos from the quantum to the universe as a whole counts on the tools available that enabled physicists to observe them. Ultimately, if behavior is what the body does, then it does not matter where the level of analysis must begin, from molecular neural events to molar behavioral events. They all equally 'count' and they all must be ultimately woven into a unified theory of behavior. A 'radical' behaviorism represents exactly this perspective (Donahoe and Palmer, 1993, Marr, 2001)

Historically, the core element of learning theories, namely reward, reinforcement or incentive motivation, has been most commonly represented by a methodological behaviorism that used as its primary subject matter directly measurable descriptions of behavior as represented by response topography or form and the contingent relationships between behavior and reward that were mapped in turn to schedules of reinforcement. This can be simply represented by act-outcome relationships that cohere with common sense appraisals of how behavior is motivated. Presently, a second factor has been added to the equation that differs not only psychologically but also in terms of the brain mechanisms that as micro-behavioral entities are also directly measurable. This second factor, formally known as Pavlovian incentive salience, is represented by the activity of the mesolimbic dopamine system in the brain, and can be more simply represented by the concept of behavioral discrepancy. Behavioral discrepancy, as postulated by Donahoe and Palmer (1993), broadens the heuristics or rules of thumb of behaviorism to encompass covert events that until only recently were unobservable. Informed by the observations derived from new methodologies (e.g. fmri, neuronal modeling) that can examine the brain in ‘action’, the common sense rules that can be applied to behavior must now not only include knowledge of the daily contingencies of cause and effect, but the daily discrepancies in the predicted correlations between cause and effect. In other words, the value of what we do is influenced not only by what we do and see, but also by the continuously changing estimate of what we predict to be.

The concept of discrepancy, or a discrepancy theory of reward (Hollerman and Schultz, 1998), simply states that the neurological sensitivity imputed to events of nativistic or inborn significance such as food, sex, power, etc. must also include abstract elements of behavior, namely the relative novelty or surprise whereby those events occur. In other words, discrepancy acts in effect as a Pavlovian unconditioned stimulus. In two factor learning theory, Pavlovian stimuli such as food, sex, etc. represented a separate learning process called respondent conditioning that complemented but did not integrate with incentive learning as reflected by operant conditioning. With the inclusion of discrepancy as a Pavlovian stimulus, this has changed. Thus not only contingency but also discrepancy must be considered in an integrated account of incentive motivation. The notion of discrepancy makes evolutionary sense, since it is unpredictable events (e.g. a predator jumping out of a tree, finding a new source of food) that have the most salience, as we need to come up with new cognitive strategies that can handle them. Thus, when we are surprised with events that occur in ways different than expected, the brain creates neuromodulators (neurochemicals that modulate or activate global areas of the brain) that fix attention, make thinking more efficient, and at high levels are perceived as having a hedonic value (i.e. they feel good). So what does this mean? It is a cause for hope, literally.



Hope

Hope, optimism, or positive thinking represents a cognitive appraisal or focus on probable or likely positive events, but not positive events that are wholly predictable. For example, I would never say that I am optimistic that the sun will rise, but for a picnic on a cloudy day, I would say I am optimistic that the sun will shine. Although optimism is a conscious appraisal of positive uncertainty, the same uncertainty and its hedonic value may be non-consciously perceived and still influence behavior (Berridge, 2001). Moreover, the hedonic value if consciously perceived may have a value that is coherent or incoherent with a rational perspective of what is good and proper.

To illustrate this, consider this mind experiment. A piece worker in a button factory may have to pull the lever on a button making machine many times during the day to make a required quota of buttons, with the reinforcement of the worker's labor occurring upon the weekly receipt of a fixed paycheck. But what if payment occurred following some random average of pulls, and what if the payment ranged from a sum many times the size of the paycheck to actual debits from the workers account? Whereas every act of the worker was before utterly predictable in terms of the relationship between behavior and reward, the sudden randomization of the size and timeliness of the reward reflects the imposition of a positive uncertainty that transforms the button machine into a slot machine. If real world examples suffice, the worker will become alert, enthused, and will likely look forward to positive results for his next day at ‘work’ even though he may recognize that his weekly winnings may on average never surpass his former and predictable paycheck of old. In other words, despite the good or bad implications of his behavior, he becomes happy and optimistic, and feels all the better because of it.

As the individual morphs from a bored and disinterested factory worker to an excited and optimistic ‘player’, the motivational source of his behavior will likely remain obscure, and his linguistic appraisal of his own behavior may thus vary radically and unpredictably, even though its neural concomitants remain as consistent and verifiable dependent measures. Thus if the worker is aware of the cause of his higher motivation, he may say that he is intrinsically motivated, and if he is aware of attendant good feelings, he may say he is also having a peak experience. Further, if the implications of his work are good, then he may refer to his experience as self-actualization, and if its implications are bad, it may be called an addiction. The worker’s behavior may therefore be conscious or non-conscious, entail good feelings or no feelings, and have good or bad implications. Each of the permutations of these factors can thus result in different mentalistic descriptions that can be assumed to reflect distinct hypothetical motivating entities that are presumably instigated by different mental or neurological processes. The unfortunate result of this is that a bestiary of inferred processes, traits, and characteristics is created that is more likely to impede than clarify an understanding of behavior. The question though remains, what is essential and what is derived? Are inferred processes the building blocks of behavior, or is there something more elemental and observable?

Intrinsic Motivators

The extrinsic vs. intrinsic motivation dichotomy in social psychology is a prime example of this difficulty. Intrinsic motivators are commonly defined as events that are reinforcing in and of themselves. That this reasoning is itself tautological, that is, ‘we behave because we behave’, and that discrete extrinsic and intrinsic motivators have no demonstrated neurological foundation has not forestalled the postulation of a host of inferred intrinsically motivating processes from ‘flow’ states to ‘senses’ of control, empowerment, etc. What is distinctive about extrinsic and intrinsic motivation states is that the former are near universally appraised as events that occur predictably and have predictable implications, while the latter are not.

For example, the intrinsically motivating flow state (Csikszentmihalyi, 1990) is a pleasurable and ecstatic state that occurs when demand matches but does not surpass skill, such as when one is performing artistic, sporting, or other creative pursuits. But demand and skill invariably co-vary, as rock climbing, tennis playing, or other creative endeavors represent continuous and unexpected shifts in the predictability of a sure grasp on a ledge, a good tennis shot, or of a moment’s inspiration. Similarly, senses of power, control, security, etc. also involve shifting estimates of the predictability of rewards, as when an individual cognitively surveys the unpredictable yet positive implications of a position of power, wealth, or other authority.

In contrast, extrinsic motivators are generally appraised as discrete events that occur predictably and have predictable implications. In particular, in educational environments, gold stars, verbal praise, and other rewards are contingent on behavior that follows predictable or rote forms, and the implications of such rewards are predictable as well. Because of the inherent predictability of such contingencies of reward as well as the reward itself, such behavior is inherently unstimulating or even boring. However, as our previous example attests, to continually vary the timeliness and size of an external reward transforms such external motivators into intrinsic motivators. Secondly, if extrinsic motivators are profuse and interdependent, a similar transformation occurs. For example, a student may find writing a play for a mere class grade a boring task to be sure, but if aspects of that story influence not only the teachers attention, but a prize from the principal, a chance of publication, the favors of a classmate of the opposite sex, beating out a competitive student, or gaining the judgement of posterity, then the uncertain dependencies between all these events would create a stimulating and internally rewarding environment. This of course was Shakespeare's predicament as well as his inspiration if the movie 'Shakespeare in Love' is to be believed.

These examples underscore the fact that the concept of intrinsic and extrinsic motivation is misconceived by assigning meaning to metaphorical (inner intrinsic and outer extrinsic) causes rather than meaning that derives from evidential characteristics of the behavior itself. That is, intrinsic and extrinsic do not denote the ‘discrepant’ value and the ‘contingent’ value that correspond respectively with the hedonic and logical ends of behavior. This analysis coheres (as it must) not only on a molar behavioral level, but also on a molecular behavioral level with present day ‘two-factor’ learning theories that divide incentive motivation into different neurological processes that separately correspond with ‘wanting’ and ‘liking’ (Berridge, 2001).

The dubious value of inferred mental motivating processes such as intrinsic motivation underscores the larger fact that human motivation has been described for millennia by a metaphorical language that is detached from or obscures the reality of behavior, and most importantly the neuro-physiological processes that make up our minds. But in biology, the homeostatic impulses or drives that animate all living things, from thirst and hunger to sex, can be dissected without primary consideration of feelings or ‘qualia’ of hunger and thirst, and do not need their introduction to be predicted and controlled. Philosophy however has always been immune to such a presumption, and has generally been considered to be not reducible to byproducts of mere physiological urges. But is this truly the case? The simple question is this: can a positive psychology be derived from behavior alone if its subjects, like its mammalian cousins are rendered in principle mute? In other words, is the necessity of virtue something that is beyond words?

The Spandrel of Virtue

In evolutionary psychology as well as popular discourse, it is presently fashionable to hypothesize mental modules engraved by evolutionary pressures that can directly account for idiosyncratic human behavior, from hungering for hot dogs to hot blondes. The evolutionary biologists Stephen Jay Gould and Richard Lewontin (1979) provided a cautionary warning to such armchair theorizing, and noted that every biological or behavioral feature does not necessarily exist for some adaptive purpose. They offered the architectural term of a spandrel as a metaphor for characteristics that are or were originally side effects and not true adaptations to the environment. Spandrels, or the gaps that surround the interstices of a series of contiguous arches, exist as a necessary outcome of building with arches. In the same way Gould and Lewontin argued that many features of organisms exist simply as the result of how an organism develops or is built. In other words, every feature does not necessarily exist for some adaptive purpose.

For evolutionary psychologists, we lust after our neighbor’s wife, and act selfishly or virtuously because of modular means-end impulses ingrained in the human brain. Similarly, inferred motivational processes, although not necessarily attributed to evolutionary selection, equally represent modular processes that occur practically independent of the brain.

But do we need inferred modules or processes to underscore human qualities, or may they be safely discarded without abandoning or denigrating their substantive predictions, namely the importance of human creativity, resilience, and moral virtue? We can if we consider such virtues to be a by-product of simple and fundamental motivational processes, and that they represent not mental modules, but spandrels.

Uncertainty is an abstract element of behavior, and any behavior that is characterized by a high, frequent, and positive uncertainty will be valued more than behavior that is elicited by events that are more predictable. Indeed, as gamblers who lose themselves to its pleasures soon note, the hopeful qualities of a die roll outweigh even putting bread on the table.

The implications of this are profound, since if value can be denoted more in the accentuation of abstract qualities of information rather than the type of information per se, then the maximization of value and the affective responses that denote that value becomes non-materialistic in nature and origin, and prospectively boundless. In turn, if happiness lies in maximizing value, then maximizing the problem sets or discrepancies or 'problems' that ultimately mediate value is the true goal of life. Thus happiness is not the mindless idyll of a land of the Lotus-eaters, but an often-painful Odyssean quest that is full of troubles. But with a sea of troubles must necessarily arise the virtues that spring from their conscious appraisal. The fact that we must be sensitive to novelty to survive impels us to be sensitive to the novelty that we not only experience in reality but the reality we must model virtually. Because we can as thinking beings model future behavior in almost infinite iterations and thus be prepared for the vicissitudes of nature, to be empathetic and to build cultures that foster empathy means to be able to model endless eventualities in nature and in human minds and to exalt the icons of music, art, sport, and literature that by their very nature entail discrepancy. Although this prepares us to survive nature by embracing its complexity, the price we must pay for this capability is to act as if virtual consequences are real, thus tying our self-interest and perhaps our very survival to shadows. The spandrel of self-less empathy, like the devil, is in the details of survival. Thus we suffer embarrassment from people who can scarcely hurt us, feel shame from behavior that creates the mere virtual approbation of others, and take prideful pleasure in acts that can never be returned in kind because we accept the currency of virtual meanings. We can gladly offer our lives for God and country because we are hopeful of the meanings they entail, and hope, a byproduct of the survival organ that is our brain, is just enough to get by.

Hypotheses Non Fingo

Hitherto we have explained the phenomenon of the heavens by the power of gravity, but have not yet assigned the cause of this power…Hitherto I have not been able to discover the cause of those properties of gravity from phenomena, and I frame no hypothesis (hypotheses non fingo), for whatever is not derived from the phenomena is to be called a hypothesis, and hypotheses, whether of occult qualities or mechanical, have no place in experimental philosophy. –Isaac Newton, Principia

In the 17th century, physics was a straightforward and settled thing. Nature, the heavens and man himself were supernaturally selected, and although obscured from view, reverse engineering nonetheless revealed the mechanics of the world. Thus, as revealed by scripture, God’s reasons for the selections He made could be divined by clerics resting on comfortable armchairs, and clever fellows like Claudius Ptolemy and Tycho Brahe who could spin earth centered mechanical universes that made input and output square. Galileo Galilei of course demurred, and cursed (vipers all!) his peers for refusing to look through his telescope. That was understandable of course, since a deep understanding of how the world actually worked was at best redundant and at worst disruptive to the conventional thinking of the time.

In the late 20th century, psychology was also a straightforward and settled thing. Nature, the heavens, and man himself were naturally selected, and although obscured from view, reverse engineering nonetheless revealed the mechanics of the world. Thus, as revealed by the geological record, evolutionary psychologists resting on comfortable armchairs could divine nature’s reasons for the selections made, and clever fellows like Tooby/Cosmides (2000) and Pinker (1997) could spin computational mental mechanics that made input and output square. Jaak Panksepp (2000) of course demurred, and cursed (autistic all!) his peers for refusing to look through the lens of neuro-psychology at the true workings of the human mind. That was understandable of course, since a deep understanding of how the brain actually worked was redundant and at worst disruptive to the conventional thinking of the time.

Presently, the influence of evolutionary and computational models of human behavior have gained great currency as explanatory devices that bypass, like astronomical models of old, a deep understanding of human motivation derived from empirical descriptions of behavior from the molecular (neural) to the molar (overt behavior). That a science of optimism, or a humanistic psychology can operate unrooted to any theory of incentive motivation or learning based upon the reality of the brain (e.g. Seligman, 2002) is one of the more astonishing and unfortunate facts of present day psychology. Until recently however, this omission could be justified by the fact that empirically rigorous neural theories of incentive motivation did not exist. Like astronomy before the telescope, learning ‘theories’ such as radical behaviorism that were based on inductive principles were impaired not by the weakness of their principles but by the unavailability of observational tools that could fully reveal the micro-behavioral facts of the human brain. Nonetheless, B. F. Skinner, like Isaac Newton, stuck to his principles, and by recognizing the danger of armchair hypothesizing, made none. But the price for this stand has been dear, namely a near universal relegation of behaviorism to the wastebasket of failed psychological principles. But this will change. The fact that bio-behavioral learning theory must have morality as its entailment brings philosophical events such as human virtue and character under the purview of empirical science. But this is hardly new even in philosophy, and traces its roots hundreds of years in the past to a Spinozan philosophy that postulated an embodied mind whose understanding entailed a just and virtuous life, and saw God, as an abstract mathematical essence, pervading all things. Ironically, the seed to this change will occur, as with the heliocentric revolution in Copernicus’s time, with a revolution in common sense.

A New Common Sense

Common sense is simple reasoning, but can radically change as long as the new paradigm that replaces it is equally simple and fits the facts of nature as we see them, aided by the instrumentalities that display them. We accept the Newtonian view of the world because of its simplicity and because it fits the facts of nature as revealed by the instrumentality of the telescope, but we do not do complex Newtonian calculations to predict where the planets will move. Similarly, a new paradigm for human behavior that incorporates contingency and discrepancy is as revolutionary as the celestial mechanics of Newton’s day, but provides us with a calculus for behavior that is even more imposing than Newton’s own.

Behavior is complex because the contingencies that instigate behavior are complex, and are cognitively denoted by a myriad interconnected and dynamic perceptions both consciously and nonconsciously perceived that are mediated by brain and body. But although we may therefore hope to generally and not exactly predict behavior, our understanding of the overriding principles of motivation need not be so uncertain. A true radical behaviorism that simply describes behavior is all that counts because ultimately all that counts is behavior, and behavior is a straightforward thing. But behavior of course is language too, and we ultimately need our metaphors to understand. Indeed, as we do not think of our pains as the mere hyperactivity of nerve endings, we will likewise always think of our lives in terms of metaphors, of hot feeling and cool reason, of flowing pleasures and walks in the park that brim with subtle surprises. A physicist uses metaphors like black holes and cosmic string so we may understand the universe, yet he must rely on higher mathematics to describe and predict. Such is the similar conundrum of a radical behaviorism that can accept metaphors for understanding, but must employ a more formal language of behavior to also describe and predict. That is why behavior is confusing, and that is why too the aliveness that is the behavior we feel and taste will forever remain a mystery. But again, we will always have hope.



References:

Berridge, K. (2001) Reward Learning: Reinforcement, Incentives, and Expectations, The Psychology of Learning and Motivation, (3), Academic Press, New York

Csikszentmihalyi, M. (1990). Flow, The Psychology of Optimal Experience. New York: Harper Collins

Donahoe, J.W. and D. C. Palmer (1993). Learning and Complex Behavior, Needham Heights, Ma: Allyn and Bacon

Gould, S. J. and R. C. Lewontin. (1979)"The Spandrels of San Marco and the Panglossian Paradigm: A Critique of the Adaptationist Programme." Proceedings of the Royal Society of London

Hollerman, J. R., and W. Schultz (1998) Dopamine neurons report an error in the temporal prediction of reward during learning, Nature Neuroscience, 1(4), 304-309

Marr, A. J. (2001) Why behaviorism, to survive and triumph, must abandon its very name. An Open Letter Behavior and Social Issues, 11(1), 92-99

Panksepp J. (2000) On preventing another century of misunderstanding: toward a psychoetheology human experience and a psychoneurology of affect. Neuropsychoanalysis, 1(2), 1-21 (on line supplement at neuro.psa.com/pank.htm)

Panksepp, J. and J. B. Panksepp (2000) The Seven sins of evolutionary psychology. Evolution and Cognition, 6(2), 108-131

Pinker, S. (1997) How the Mind Works. Norton: New York

Seligman, M. (2002) Authentic Happiness. Free Press: New York

Tooby, J. & Cosmides, L. (2000) Towards mapping the evolved functional organization of mind and brain. In: Gazzaniga, M. S. (ed.) The New Cognitive Neuroscience (2nd edition). MIT Press: Cambridge, Ma., 1167-1178

No comments: