Stack Overflow recently published the results of its Developer Survey. It came at a perfect time: we had just discussed the difference between a Software Engineer and a Data Engineer at Automattic and the survey results about Correlated Technologies demonstrated the difference in tech stack for these two roles.
They looked at the technologies respondents use and drew a network graph by connecting two technologies together if they are used by the same person. I feel like it gives a good picture of the day-to-day tech stack of our data scientists, and data and software engineers — though, of course it isn’t the full picture. For example, some of our engineers use Elasticsearch.
Other sections that caught my attention were around ethics in technology. Our team is quite interested in bias and fairness in tech and especially AI; we read and talk about it. I believe it is a topic that every data scientist will eventually have to make a conscious effort to deal with. So it makes sense that data scientists were at the top of the list when it comes to how much more likely they are to to be concerned about fairness over singularity when it comes to AI (I can’t even fathom why anyone would be more concerned about science fiction than built-in discriminatory biases, but that is another topic).
I was happy to read that 79.6% of respondents believe that developers have an obligation to consider the ethical implications of their code. Of course, when talking about ethics, the question of responsibility comes into play as well. Interestingly, when the question is “Who Is Ultimately Most Responsible for Code That Accomplishes Something Unethical?” only 19.7% of the respondents said “The developer who wrote it,” while 57.5% responded “Upper management at the company.”
However, in the AI context, according to the respondents that responsibility seems to fall more in the executioner’s hands:
That some responded “Nobody” makes me think that some people might have interpreted this question as “Who is currently responsible for considering the ramifications of AI?” rather than “Who should be responsible?” as it can be interpreted both ways (which I think is a survey design flaw). It would be interesting to see the responses to these two questions about ethical ramifications with both of them having the same options available for answers.
The survey is highly biased toward young, straight, white men with less than five years of professional experience, who don’t have children. Some of the answers get really interesting when we look at the results segmented for example, by gender.
My favourite one is the differences in assessing jobs by gender:
Different types of developers apply different sets of priorities when considering jobs. Developers who belong to gender minorities in tech rank the company culture and office environment as their highest concern when assessing a new job. The gender identification question allowed respondents to select all that apply.
The majority of non-male respondents picked company culture and office environment as their highest concern, while the majority of men picked compensation and salary. I found this really interesting. My issue is that the results only show the percentage of people who picked the given item as their top priority. I would be more interested in seeing a weighted ranking.
Also, it led to some annoying online discussions by people who interpreted the results to mean that for example only 4.3% of women respondents care at all about diversity — which is not what the results are saying, but the framing is misleading if you don’t read the descriptions (which we should).
All in all, the survey is an interesting read, I am really glad that Stack Overflow is putting in the time and resources to carry them out. I hope they will make the raw data available like they have in the past!
Images in the body of this post are sourced from the 2018 Stack Overflow Developer Survey.
Data for Breakfast 