TITLE: Student Learning as Confronting Risk AUTHOR: Eugene Wallingford DATE: May 10, 2007 11:28 AM DESC: ----- BODY: Last time, I wrote about how some ideas on human psychology, from Bruce Schneier's The Psychology of Security paper. Part way through, Schneier jokes:
(If you read enough of these studies, you'll quickly notice two things. One, college students are the most common test subject. And two, any necessary props are most commonly purchased from a college bookstore.)
University psychology researchers are as lazy as university computer scientists, I guess. Schneier doesn't mention a question that seems obvious to me: Does this common test audience create a bias in the results produced? If college students are not representative, then the results from these studies may not tell us much about other kinds of peoples' behaviors! In many ways, college students are not representative of the rest of the world. They are at a nexus in development, different from teenagers at home but typically not yet living under the same set of constraints as people in the working world. But I'm not too worried. Enough other studies on risk and probabilistic reasoning have been done with adult subjects, and they give similar results. That isn't too surprising, because what we are testing here doesn't involve reflective choices that are conditioned by development or culture, but rather reactions to conditions. These reactions are largely reflexive, under the control of the amygdala, "a very primitive part of the brain that ... sits right above the brainstem.... [It] is responsible for processing base emotions that come from sensory inputs, like anger, avoidance, defensiveness, and fear." But overthinking Schneier's joke got me to thinking something else: how do these ideas apply to students in their own world, where they score points for grades in a course, make choices about what they need to know, and incur the costs of studying something now or later? Prospect theory tells us that people prefer sure gains to potential gains, and potential losses to sure losses. I often observe students exhibit two behaviors consistent with these biases: There may be simpler emotional explanations for these behaviors, but I am thinking about them in a way way in light of Schneier's article. Like software developers in general, students certainly fall victim to optimism bias. I've always figured that, when students gamble on getting more work done than they reasonably can in a short period, they were reacting to a world of scarce resources, especially time. But now I see that whatever conscious choice they make in this regard is reinforced or even precipitated by a primitive bias toward optimism. Their world of scarce resources is in many ways much like the conditions under which this bias evolved. Further, this bias is probably reinforced by the fact that college CS students are just the sort who have been successful at playing the game this way for many years, and who have avoided the train wrecks that plagued lesser students in high school or in less challenging majors. It must be a shock to have a well-trained optimism bias and then run into something like call/cc. Suddenly, the glass isn't half full after all. A few posts back. I wrote about reliance on external references. For students, I think that this turns out to be dangerous for another reason, related to another tendency that Schneier documents: the availability heuristic. This refers to the tendency that humans have to "assess the frequency of a class or the probability of an event by the ease with which instances ... can be brought to mind". People are overly influenced by vivid, memorable instances. When they have encountered only a few instances in the first place, I think they are also overly influenced by the instances in that small set. An instructor can go to great lengths to expose students to representative exemplars, but that small set will also have the potential to mislead when relied on too heavily. Relying on external references to recall syntax is one thing; it will usually work out just fine, even if its is unacceptably slow, especially in the context of an exam. But relying on triggers for more general problem solving can create problems all its own... The most vivid, most memorable, or only instances you've seen will bias your results. I am a strong proponent of reasonable from examples, a lá case-based reasoning, but this requires a disciplined use of a reliable "similarity metric". Students often don't have a reliable enough similarity metric in hand, and they often haven't learned yet to use it in a disciplined way. They tend to select the past example that they remember -- or understand!! -- the best, regardless of how well it applies in the current context. The result is often a not-so-good solution and a disillusioned student. Thinking these thoughts will help me teach better. -----