Tuesday, July 2, 2013

Analogy Works, Linearity Doesn't: a tribute to Leslie T. Wilkins



IN HUMAN AFFAIRS: ANALOGY WORKS, LINEARITY DOESN’T
A Tribute to the Wisdom of Leslie T. Wilkins

Hal Pepinsky, pepinsky@indiana.edu, “peacemaking” at Pepinsky.blogspot.com
July 2, 2013

                 This blog post follows on my last, on “ordination addiction.”  It is an argument for learning the social consequences of our interactions at all levels  by the stories we live and share, for leaving linearity in its place, how to build physical tools.
                I have been guided by the statistical wisdom of Leslie T. Wilkins, since he invited me to team teach a seminar with him and a rotating philosopher of science the first time as a new faculty member in the fall of 1972 I went to his office.  Armed with a 2-year engineering degree he got pioneering application of social systems analysis to reducing the risk of plane crashes.  The question for him became, “Do you replace the pilot or reposition the instruments?”  as a member of the Home Office Research Unit won the Royal Statistical Society Award for predicting which juveniles would recidivate in 1954, then worked for the UN and politically incorrectly served as dean of Berkeley’s school of criminology, until Dick Myren hired him at Albany in 1968.  My first day of work I made a point of visiting his office.  He showed me the proofs of his 2-paragraph foreword to Wenk and Emrich’s impeccably designed study  from  Les’s point of view yet failed attempt to predict which “violent” California youth would recidivate (in the Journal of Research in Crime and Delinquency, 1972).  His conclusion, “This ought to be the last word on prediction.”   Les, Barb, Jill and I kept in touch continually for a decade after I left Albany in 1976.  I especially remember the time that Jill, Katy and I stayed with him perhaps in 1986 and Barb at home in a cozy old police station/jail.
                Les taught me 2 axioms that have guided all my social analyses since they sank in.
How is more important than why.
Stochastic models are the only social models that work.
                Les and Don Gottfredson collaborated in designing the eventually notorious US federal parole guidelines.  The system Les designed assumed that since experimentally parole board members had settled on a regression model of their decisions that accounted for some ¾ of their decisions, he could expect them to depart from the guidelines in at least 15 percent of cases.  The system required that a board member write reasons for every departure from the guidelines.  The board would review the reasons annually with a view to revising the guidelines to accommodate their exceptions (in statistical terms, their prediction errors).  Les and I never hesitated to disagree with one another.  Whenever he mentioned his plan, I would tell him I expected that board members would use the guidelines as a bureaucratic tool for absolving themselves from responsibility for the rightness or wrongness of their decisions.   We were predicting in the same stochastic frame.  He thought the decision-making machine would drive the board.  I figured that they would behave like any politically visible and vulnerable body, and love to be able to say, “I was just following the rules.”  Within stochastic assumptions, the probability that either of us would prove right was unknowable, and both of us knew that well enough not to press the point much.  Les proved wrong, and I didn’t anticipate the extent to which the Board and Congress would revise the guidelines to suit political purposes.  It happened at the time that I was still deep into trying to understand Mao’s rejection of the rule of law.  In China I saw a recognition that lawful rights ultimately defended private property rights, where owning trumps not-owning, even in daily “contractual” life.  The Nuremberg decision that following orders is no excuse became a source of national moral pride during my post-WWII childhood.  Ultimately, rules of law bent to politically prevailing winds.  As with the parole guidelines and their successor guidelines, the force of political convenience, as in the habit of appearing to follow “the rules,” runs strong, as by “zero tolerance” for those caught “breaking the rules.”
                Enduring fables become memorable because they at once are not strictly deducible and yet resonate with our own experiences of what we get for what we have done.  They resonate because the same sequence of action-reaction keeps getting the same results.
                The trouble is, that no one can “prove” that an analogy is true or false.   Reactions to our actions transcend linearity.  On one hand, we each other it is foolish to repeat history and expect different results.  On the other hand, when I have tried to analogize past events to current events, it historians have been prominent among those who have admonished me for overlooking uniquenesses any discrete past event.   I am reminded me of law school training, where we were taught the skill of “I can distinguish your case” in briefing any side of any legal dispute.  It is by contrast an act of faith to act on the hypothesis that the reaction to what we do will be like those other reactions, most of all when what we try deviates from standard practice--what now in social sciences are heralded as “evidence-based best practices,” and from what our elders and keepers tell us are “good decisions.”  It is only when events disappoint our expectations that we can learn from history.  It is only when “their” reactions recapitulate “our” reactions that we have an empirical ground to distinguish what works (get what we want) from what does not.  This to me is the implicit message underlying the golden rule; it works more often and sustainably than doing to others what “they” need or deserve that “we” don’t.  Thomas Kuhn in The Structure of Scientific Revolutions observed that the point at which people decided that they had reached a point where an entire “field” of “scientists” had reached a point the history of disappointed expectations would be overcome by enough “scientists” happening amidst failed attempts to come upon an analogical frame for selecting and processing information that worked better—a new “paradigm.”  At a personal level as I have observed at “higher” social levels, I have found that a whole range of ways I have tried to gain and share social security with other that do not work, and amidst the rubble, some that keep on making my own life more socially secure, let alone the lives of disputants among whom I have tried mediating rather than disposing of differences.
                Regardless of whether I really know anything about what I and others can expect to get back from others what we intend to make them do, I see no logic whatsoever in projecting results of things we do to others now, from what they or others “like them” did when left to their own devices.  We may be able to confine people so completely that we can even force them to eat or lie still.  We may be able to stop people doing what we don’t want to do by killing them, but not to stop their survivors and kindred spirits from retaliating or escaping our political grasp.  When we move ourselves and others from point A to point B, no statistic can foretell which direction movement to point C will take in any individual case.  No logic of empiricism justifies using information about what people are, to decide what they deserve or need to get.  Stories of how people reacted in circumstances that seem like mine or ours are the only social information that has ever worked for me.
                Les Wilkins used to talk with a twinkle in his eyes of assigning his advanced research methods students to go out and find serendipity.  I have tried it, and it has worked for me as it did for him.  Thank you, Les, for lending me your world of human understanding.  Love and peace--hal

No comments:

Post a Comment