Is there a Geek Gene? Are CS grades bi-modal? Moving computing ed research forward
January 21, 2020 at 7:00 am 14 comments
This month’s Communications of the ACM published Elizabeth Patitsas’s ICER paper about bimodality in CS grades (or rather, the lack thereof) as a research highlight, Evidence that Computer Science Grades are not Bimodal. It’s a big deal to have computing education in this position in the ACM flagship publication, and thanks to Shriram Krishnamurthi for his efforts in making this happen.
I wrote about Elizabeth’s paper when it was originally published at ICER at this blog post. Elizabeth wrote a guest blog post here on these topics (see here). These are important issues — Wired has just published an article talking about the Geek Gene with a great discussion of Betsy DiSalvo’s work (see post here about some of Betsy’s work).
I wrote the introductory page to the article (available here). I point out that Elizabeth’s article doesn’t end the debate, but it does move forward how we address questions about how we teach and how students learn:
This paper does not prove there is no Geek Gene. There may actually be bimodality in CS grades at some (or even many) institutions. What this paper does admirably is to use empirical methods to question some of our long-held (but possibly mistaken) beliefs about CS education. Through papers like these, we will learn to measure and improve computing education, by moving it from folk wisdom to evidence-based decision-making.
Entry filed under: Uncategorized. Tags: computing education research, Geek Gene, undergraduate education.
14 Comments Add your own
Leave a comment
Trackback this post | Subscribe to the comments via RSS Feed
Computing Ed Research - Guzdial's Take 
1.
alfredtwo | January 21, 2020 at 9:50 am
Some students seem to get it right away and some students never seem to get it. Even if there is no bi-modality there is a range of scores in any class. The same is true in any subject though. How much of that is the teacher and how much is the way a student;s brain works?
2.
Mark Guzdial | January 21, 2020 at 10:23 am
“Never seem to get it” — I think the critical word here is “seem.” There is an economics argument that says that the amount of time necessary for an under-prepared student is too great to be effective (see post here). But all the research evidence I know suggests that that’s not true — everyone gets better with practice.
3.
alanone1 | January 21, 2020 at 10:49 am
I don’t think it’s a good idea to use pop culture terms and tropes — “geek”, and “there’s a gene for X” — when trying to do something that aspires to science.
We can say “Biology is variation”, and back it up with overwhelming evidence. This will imply that a random collection in any species with show differences, and with some traits being much more rare than others.
Most creatures learn some parts of their behaviors, and we can now substitute “ability” as a non-linear combination of “inborn”, “acquired skill”, and “will”. This will also be a distribution.
One of the goals of formal education has been to move the abilities of as much of a population as possible to get above important thresholds of fluency. If this is done well, we would expect to see very few below threshold individuals, lots above threshold, and some very unusual individuals at the high end. For example, this is what we’ve seen in the past highly literate Japanese society.
A question to ask here might be about music. Is it bimodal? It might test out that way, but very different approaches to teaching music in some towns have shown something similar to Japanese literacy: an entire population of children can be brought to fluency (and there will also be “unusual abilities”).
In our work, we decided in the 70s that our pedagogy was failing if we couldn’t do the same with children and computing. This set us up for almost 25 years of failure, but we finally found/devised a combination of approaches, resources, and teaching that achieved this.
Since skill is being asked in many cases to make up for “inborn”, and in these cases “will” has a huge affect on the amount and kind of effort put into gaining skills, it’s worth trying get estimates of some of the things that will be required.
A good rule of thumb that we used in our decades long work with children is “pick 5”. (This is an arbitrary number larger than 4.) So: what kinds of help can we give? Some children will hardly need help, there are perhaps 3 kinds of help we can come up with, and some children we will not know how to help.
And: what kinds of motivation will we see and can we inspire? Some children will be highly motivated, we can probably figure out 3 different ways to appeal to different types of children. And there will be a few that we won’t be able to motivate.
It is very often the case that children who don’t need a lot of help will be motivated by their early successes. (Etc.)
Our composite rule of thumb turned out to be that we might have 16 to 20 combinations to deal with. This is one way to see why Montessori’s ideas about embedding schooling into the cultural fabric of a school is so important. The drive to acquire the surrounding culture is an Ur-motivator, and will drive many children to make up with skill what they might lack at birth.
Of course, the above is still too simple, but it is much better than getting trapped with bad/wrong metaphors that don’t obtain to the way genetics and learning really work.
4.
Mark Guzdial | January 21, 2020 at 11:01 am
So nice to hear from you, Alan! I agree with you. Elizabeth’s result isn’t about genetics at all. Rather, she deals with teacher opinions and judgements, and with actual grade distributions. We’re still learning to do science in computing education (e.g., we’re still working to develop good measures, because it’s hard to do science without them). I see Elizabeth’s paper as a positive move in that direction.
5.
alanone1 | January 25, 2020 at 12:55 am
Hi Mark
The opinions of a population are sometimes (often?) bimodal, but I’m straining to identify genetically generated distributions of cognitive traits that are. But as we move from the “inborn” to the “abilities” composite, it’s easy to see how you can get bimodality rather than a smoother distribution. For example, the environment and past histories can create big differences in motivation and will that are key factors in developing skills.
Another key principle (I think) is that it is in poorly set up environments — especially socially, especially in the culture around the learners — in which you will see bimodality. Under these conditions one would expect that “inborn” predispositions will have a much bigger effect on motivation and will.
6.
gasstationwithoutpumps | January 25, 2020 at 12:22 pm
One can turn any distribution of students into a bimodal distribution of scores simply by testing exactly the same thing over and over and adding the scores.
Similarly one can get a normal distribution by adding a large number of independent tests.
The distribution of scores often says more about the measuring instrument than about the population being measured.
7.
alanone1 | January 25, 2020 at 12:25 pm
Yes, your last sentence is a very important point.
8.
Don Davis | January 25, 2020 at 1:48 pm
I forget the term at the moment – basically, educationese for “tipping point.” Might a bi-modal distribution represent a sort of tipping point? [With the difference to a “geek gene” being proficiency with an underlying skill (cf. one / some of the CT proficiencies.]
9.
alanone1 | January 26, 2020 at 12:43 pm
Another perspective on all this is to look at activities where small differences translate into large outcomes.
For example, tennis is a “balance sport”: what you can do and how well depends ultimately on whether you are on or off balance. An opponent who is only slightly better/faster can win with lopsided scores because if they can beat you to the punch they can put you off balance and this will either result in a out ball or a weak return, and the next shot from the opponent can put you in worse shape, and ultimately you will usually lose the point.
In music, a slightly better memory — especially muscle memory and “sight memory” — will translate into big differences in many areas, and especially how much ground can be covered.
There is something similar in math (called “projection”) which is what kinds of things pop up when looked at relationships.
In each of these cases being just slightly behind the eight-ball feels like way behind the eight-ball, and can be very discouraging. Tim Gallwey’s “Inner Game” methods are very good at getting to what can be started with and improved.
For the learning of most things we shouldn’t expect bi-modality if the processes are set up well.
On the other hand we can instantly get something that looks like bi-modality by upping the ante and drawing a very high threshold. This will make it extremely more difficult for any given person to get above the threshold by skill learning alone.
10.
bigdlovin | January 26, 2020 at 3:25 pm
What an excellent reply. I suspect I was thinking of “threshhold concept” – but more of a threshold skill or behavior. Working with high school students, especially those with low thresholds for frustration, I see how that could translate to all or nothing for some of them.
I am not sure if this corresponds to what you meant with “more difficult for any given person to get above the threshold by skill learning alone.” It’s what I think of because looking back I was only successful with early programming because of ‘ persistence’ (or delayed schedules of reinforcement or whatever you want to call it). I didn’t really “get” the core abstractions of programming until I “got” objects. [Of course, really “getting it” is relative.] I just kept plodding ahead.
11.
Tom Morley | January 24, 2020 at 5:38 pm
Sometimes beginning calculus is bimodal, but I expect this is due to big differences in high school algebra.
12.
Changing Computer Science Education to eliminate structural inequities and in response to a pandemic: Starting a Four Part Series | Computing Education Research Blog | July 13, 2020 at 7:00 am
[…] (UMdCS). We teach in ways that assume a fixed mindset — we presume that some students have a “Geek Gene” and there’s nothing much that teaching can do to change that. We know the opposite — […]
13.
Proposal #2 to Change CS Education to Reduce Inequity: Make the highest grades achievable by all students | Computing Education Research Blog | July 27, 2020 at 7:00 am
[…] mindset. If you believe that there is a random distribution of “talent” or “ability” or “Geek Gene” in the course, and (critically), there’s nothing much that teaching can do to change that, […]
14.
Dijkstra’s Truths about Computing Education Aren’t: The many kinds of programming | Computing Education Research Blog | December 7, 2020 at 7:01 am
[…] with programming” (see interview here). We have a lot of evidence that there is no Geek Gene (see this blog post here), i.e., we have note yet identified innate differences that prevent someone from learning to […]