this is a common and inherent phenomenon in CS classes.
The perception of bimodal grades provides evidence to the
Geek Gene narrative that some students “have it” and
some do not. And when new educators who have been
primed to see bimodality then begin teaching and do not
see all their students learning, these new educators can
then see this as evidence of the Geek Gene. The reproduction of the Geek Gene Hypothesis is hence social in nature.
5. 2. The “Geek Gene” is an equity issue
Debunking the “Geek Gene” is also important for equity
reasons. Recent studies have found that academic disciplines in which “brilliance” is seen as necessary for success
have less gender diversity.
13 Looking at the history of science, women and people of color were long denied entry
and acknowledgment in science because they were seen
as lacking the “brilliance” needed to do science.
23 If computing ability is viewed as being the result of a “Geek
Gene,” then educators may use this as a reason not to
teach students who they perceive as lacking this “gene.”
Similarly, they could lower expectations of these
groups and encourage them less—which is troubling
given evidence that teacher expectations have an effect on
Our analysis of one institution’s CS grades indicates that
although bimodal grade distributions can be found, they
are far from typical. Much more commonly, grade distributions are normal ( 85.1% of cases) or highly skewed unimodal distributions. Our psychology experiment found
that participants who were more likely to label ambiguous distributions as bimodal were also more likely to
report a belief in an innate ability to succeed in CS. This
suggests that instructor beliefs play a role in the perception of bimodality.
Priming participants to think about the common perception of bimodal grades also led to participants being
more likely to label ambiguous distributions as bimodal.
This suggests that confirmation bias plays a role in the
belief that bimodal grades are typical.
Given that the belief that CS ability is innate is widespread among CS educators, there is likely a social element to the confirmation bias. This belief in bimodality
appears related to the belief in innate ability, which in
turn has been implicated in the under-representation of
women and minorities in computing. We encourage educators reading this paper to take time to analyze the
grades in their own classes, and bring the same level of
rigor and skepticism we would use in our research to
understand our own teaching.
The first author received funding from the Social Science and
Humanities Research Council of Canada. We would also like
to thank our anonymous reviews, Aditya Bhargava, Jinghui
Cheng, Jeff Forbes, Jin Guo, Mark Guzdial, Ray Lister, Andrew
Petersen, Greg Wilson, and Dan Zingaro for their feedback
and suggestions on this line of investigation.
Furthermore, belief can affect judgment regardless of
ambiguity. For example, Kahan et al. found that participants were more likely to get a math problem incorrect if
the correct result would disagree with their political
12 It is hence plausible that a computer scientist
who believes in the Geek Gene Hypothesis could look at
an unambiguously unimodal distribution and still view it
5. THE GEEK GENE HYPOTHESIS
AS A SOCIAL DEFENSE
Once again, our findings support Lister’s hypothesis that
CS grades are generally not bimodal and this perception
stems from instructors expecting to find bimodal grades
due to a belief in the Geek Gene Hypothesis. We now go a
step further and argue that the perception of bimodality
is a social defense in the CS education community.
In sociology and social psychology, a “social defense is
a set of organizational arrangements, including structures, work routines, and narratives, that functions to
protect members from having to confront disturbing
emotions stemming from internal psychological conflicts
produced by the nature of the work”.
5. 1. Social defenses in teaching
Guzdial reports that teachers generally have a high level of
self-efficacy (great confidence in their teaching ability) at the
start of their career. This then plummets as they face the
realities of classroom teaching but slowly returns with time.
Teacher self-efficacy is not necessarily tied to teaching ability: university educators often get little meaningful feedback
on how their students are learning, given their large class
sizes and lecture-based pedagogies.
Guzdial notes that if an individual university-level CS
educator has high self-efficacy, and sees evidence of students not learning, then it is rational for them to believe that
the problem lies with the students and that the problem is
innate to them—that is, beyond the ability of the teacher to
9 Compounding this, Sahami and Piech have
observed that CS educators are more aware of their top and
bottom students than they are of their average students, giving educators a biased perception of their students’ abilities.
24 Guzdial argues that CS educators have poor results,
because we so frequently use ineffective teaching methods.
Zingaro et al. suggest that not only do CS educators frequently use ineffective pedagogies, they also frequently use
ineffective assessment tools.
We theorize that the Geek Gene Hypothesis is a social
defense: it is easier for computer science educators to blame
innate qualities of their students for a lack of learning than
it is for the educators to come to terms with the ineffectiveness of their teaching.
A social defense is a phenomenon on a social scale, in
contrast to Guzdial’s observation about individual teachers. When numerous educators bond over how their students just “do not have it,” it allows for the Geek Gene
hypothesis to go from one individual’s suspicion to a
social narrative. And as bimodal grade distributions
sometimes do occur, those cases are used to argue that