How Can We Promote Valid Learning?
Ways to overcome the pitfalls of learning in a one-shot experience.
Posted Jan 19, 2018
In the movie Groundhog Day (if you haven’t seen it), Bill Murray plays a narcissistic TV weatherman, who is sent to Punxsutawney, PA, to report whether the groundhog, Punxsutawney Phil, sees his shadow. Murray, instead, finds himself repeating the same day, again and again. It doesn’t matter what he does, he’ll wake up again on Groundhog Day, in the same bed and breakfast, with the alarm clock radio waking him up to the song, “I Got You, Babe.” Recognizing his predicament, Murray begins to indulge in every base-impulse—leading to humorous scenes of Murray gorging on every donut in a diner, and driving Punxsutawney Phil off a cliff in a pick-up truck, among other sordid examples.
Throughout the movie, Murray learns from experience, that his selfish and indulgent behavior leads to repeated negative outcomes. This includes being repulsive to his love interest (Andie MacDowell). Slowly, over dozens of cycles, his character begins to transform into a “Renaissance Man.” He learns to speak French and how to play the piano. Humorously, his helpful actions around town make Murray the target of a bidding war at a charity bachelor auction that evening. Needless to say, Andie MacDowell can’t help but be impressed at Murray’s redemptive transformation.
Contrast how Bill Murray learned from experience with one-shot experiences. When we are in one-shot experiences (e.g. choosing a college, a first job, a home) it is much costlier to “learn from experience” (see also Thaler, 2015, Chapter 6). In one-shot experiences, we can’t alter our approach slightly and see the results we get (as Bill Murray did), it is harder to learn what will produce better outcomes.
Without many feedback cycles to validly learn from experience, we are more prone to “superstitious learning” (March & Olsen, 1975), meaning in the search for causality, we attribute a prior act as the cause of what we experience. Maybe we received a “lucky rabbit’s foot” as a gift prior to making a big, one-shot decision. We might superstitiously attribute the consequences we experience to that prior act. Or, more relatedly, we might “learn” that the shadow of a groundhog somehow predicts future weather conditions.
Likewise, when we try to explain organizational performance, we may attribute some social norm as the cause of performance. This may be plausible, but in other cases, it is an extraneous social norm that has no causal impact. For example, I often teach a case study about Apple and the turnaround of the company by Steve Jobs. Steve Jobs had many strengths and talents, but he also had some well-known tendencies to treat people in a less than respectful manner. It’s cognitively-tempting to attribute the success of Apple to the accountability created by his berating-style of management. We’ll superstitiously learn that a berating management style leads to superior performance. In the one-shot experience of Steve Jobs turning around Apple, we cannot go back in time and see if they still would have succeeded had they removed Jobs’ style of management. Perhaps it was primarily Steve Jobs’ unique eye for talent and his focus on the beauty and simplicity of design that was causally important, and Apple succeeded in spite of his berating management style.
The point is to recognize if we are learning like Bill Murray in a Groundhog Day-like environment where we can improve with constant feedback and repetition (see also Kahneman & Klein, 2009), or if we are in a one-shot experience. If we are in a one-shot experience the likelihood of superstitious learning is high—heightening the need for tempering our conclusions. This also heightens the need to learn how to learn. By this, I mean more systematically designing experiments to foster more valid learning. For example, rather than implementing only one option (given it is cost effective), we can test two options (much like A/B testing in web analytics where users receive one of two options and the designer can see which option leads to higher click-through rates.) By testing two options we can more validly learn what produces better results, rather than making attributions that may be superstitiously-derived.
So the next time you see yourself attributing causality (and more importantly taking action based on that attribution), ask yourself if there are other plausible attributions and how you know what you know. And, given the situation affords the opportunity to test various options, work to design ways to more deliberately create feedback cycles to validly learn. You may not redemptively transform like Bill Murray does, but you’ll increase the odds of reaching your goals.
Kahneman, D., & Klein, G. (2009). Conditions for intuitive expertise: A failure to disagree. American Psychologist, 64(6), 515-526.
March, J. G. & Olsen, J. P. (1975). The uncertainty of the past: Organizational learning under ambiguity. European Journal of Political Research, 3,147-171.
Thaler, R. H. (2015). Misbehaving: The history of behavioral economics. New York: Norton