Using mentoring to increase academic performance
Helena Horvat UA’23 and Jim Stellar
We recently wrote a blog post about how mentoring might develop between a student and a faculty member and how that could be productive in terms of a student’s career planning. Now we want to turn our attention to how mentoring might improve the academic performance in college
To do that, we found an old summary study by Bloom (1984) who looked at how strong tutoring (i.e. in-class mentoring around content) produced a large improvement in graded performance. The results are impressive enough that we reproduce below the central figure from that study. Before explaining the figure, we also note that the topic was a central point early in a Ted talk by the founder of the Khan Academy, focusing on ChatGPT as a tutor (to which we will return).
Notice in the figure above, the curve in the middle that is labeled “conventional.” It is a normal distribution – the so-called bell curve. It is the grade distribution that teachers are supposed to get in their exams with an “A” grade being at the high end, to the right of the conventional curve and with most of the scores in the middle.Then notice the curve labeled “tutorial” on the right. This is basically what Bloom found after the application of good tutoring to each student individually on the same subject matter. The effect is to shift the results so that the average student in the tutorial condition has a grade that would have been exceptional or an “A” grade in the normal distribution.
Bloom called this shift the “2-sigma problem” because it is statistically a 2 standard deviation shift where the standard deviation is called “sigma.” It places the average of tutorial distribution at the 96% level of the conventional distribution. Bloom celebrated how great it would be if we could make our average students today into “A” students in the future. Bloom called it a problem because he had no idea how in a high school or college class of normal size, a teacher could have the time to produce the high-quality and individual tutoring that is required.
We do not have a solution for this scaling problem. But we already mentioned above a TED talk by Sal Khan where he pointed to his adaptation of ChatGPT as a possible artificial intelligence (AI) solution for an academic tutor that can scale. The hope expressed in that Ted talk is that AI tutoring, like mentoring for general student development, can help the student to have the confidence, the motivation, the energy to be their own knowledge tutor with his version of ChatGPT. Of course, that remains to be seen, but the idea is intriguing as AI scales to potentially solve the teacher’s time problem.
Back to mentoring for academic content learning
As we mentioned before, the result of a mentoring bond between a faculty member and a student is often a jump-start in the student’s behavior towards meaningful goals. What mentoring and tutoring have in common is building student confidence, opening imagination, and generating effort applied to the task. We see this process as almost like a dopamine stimulant applied to the limbic system that energizes harder deeper academic work. However, like tutoring above, the current form of faculty-student mentoring does not scale. It too takes time and depends upon forming a bond between a mentor and a mentee.
While a student may comprehend concepts taught in class or be tutored by AI, we believe that the in-person interaction combined with the symbiotic relationship of mentoring further reinforces learning and growth. The closeness of a mentoring relationship keeps the mentee from getting lost in the ivory tower crowd of college students and naturally assuming one is not capable of greater work. Emotional rapport and genuine interest in one another open doors for combining the logic of specific steps towards a career goal as well as unpacking experiences where the student’s enthusiasm and joy are awakened (or not). The key is understanding what made something exciting allows you to seek out similar opportunities, and utilize your strengths. Discovering what made something unpleasant can be also useful with a mentor because they can work with you as a student to reaffirm your doubts or inform you that your particular experience was rare, or develop another focus.
Often when we reflect on the past, our strongest memories are charged with emotion. This is why the human quality of one-on-one mentoring is unmatched compared to the more sterile lecture-based classroom instruction or the individual online career search. Someone taking a personal interest in you leads you to believe you are worthy of investing time in and gives you a structure to follow. Set meeting times with a tutor or mentor holds you accountable because you are making the promise to show up not just to yourself, but to another person as well. Additionally, a mentor helps you find greater meaning in your work by nudging you to future goals that integrate your strengths. It is not clear that Khan’s content tutor using ChatGPT would work in place of career development mentoring even if it worked in a classroom knowledge-content setting.
Again with the brain and experiential learning
Given that this blog series has a focus on the brain-basis of the power of learning from experience, and given that we believe that mentoring, or even tutoring, can and should rely on emotional connections, it has much in common with experiential education. Due to the power of emotion integrated with cognition in learning from experience and in making career-path decisions, this type of tutoring also suffers from the same problem of integrating cognitive-conscious-neocortical planning with emotional-limbic-evaluation. Namely, our brains do not seem to be built so that we make this integration easily – hence the quote we often cite by Pascal “The heart has reason of which reason does not know.” And hence the need for reflection; a key component of experiential education, as well as mentoring and good tutoring.
In the Khan TED talk about AI tutoring referenced above, he makes a point out of saying that they have gotten ChatGPT to not just give one the answer, but to work with the student to help them find the answer. At one point, he describes it as almost “magical.” Perhaps that is what it takes for AI to be the best tutor and not just a fancy Google look-up device. Two questions then arise: 1) Can this AI tutoring really do that now or in the near future? and 2) Will universities that are filled with people avoid becoming an AI ivory tower where a student might again feel lost?
