operant conditioning

How Can Behavior Analysis Assist in the Study of Machine Learning?


Recently, we have been contemplating the relation between machine learning and behavior analysis.  Interestingly enough, this topic of the collaboration of machine learning and behavior analysis is not new.  William Huchison, in 1998, discussed "Computer Simulations of Verbal Behavior," and compared the learning of verbal behavior in computer simulations to that of a human learning verbal behavior, proving that "operant learning is sufficient to produce verbal behavior."  Fast forward to 2012, and in Huchison's work, "The Central Role for Behavior Analysis in Modern Robotics, and Vice Versa," it is observed that most computer "systems assume basic foundations of operant conditioning (Huchison, 2012)."

Howard Rachlin (2012) also discusses the similarities between the IBM computer, Watson, and a human being, in his work, "Making IBM's Computer, Watson, Human," outlining that the basic needs of humans and computers are the same:  we need an energy supply, protection from the elements, maintenance, and our bodies to be maintained at a reasonable temperature.  He theorizes that by creating a "Watson II" with more characteristics of human beings (i.e. being able to "override its own logical mechanism" after establishing "rules"), ultimately:

"those who would see Watson II as nonhuman because he was manufactured, not born, might go on to say that it would be worse for humanity were we all to be made as Watson II may be made. To me, this would be a step too far. We are all a conglomeration of built-in and environmentally modified mechanisms anyway. And no one can deny that there are flaws in our current construction." 

In both of these articles, Huchison and Rachlin, create a call to action, discussing the need for collaboration in computer science and behavior analysis, as computer science evolves to machine learning and artificial intelligence.

So how do we do this?  Ultimately, computer science majors interested in programming, should receive training in behavior analysis.  Currently, this is not happening, as most computer science programs have adopted more of a "pop psychology" background in their current research and practice (see Ink's article:  "MIT Researchers Use Reddit to Create World's First Psychopath AI).  What this further demonstrates, is not only a need for the dissemination and training of our science to the computer world, but also a need for a similar ethics code in the shaping of behavior when programming, as we do as behavior analysts working with human beings.

So what do you think? Comment below!



Hutchison W.R. (1998). Computer simulations of verbal behavior for research and persuasion. The Analysis of Verbal Behavior. 15:117–120. 

Hutchison, W. R. (2012). The Central Role for Behavior Analysis in Modern Robotics, and Vice Versa. The Behavior Analyst35(1), 29–35.

Rachlin H (2012).  Making IBM's computer, Watson, human. The Behavior Analyst. 35:1–16