NEW RESEARCH PAPER OUT! In this paper, @grimalkina and I explore how #code review anxiety is maintained and exacerbated to develop a model of code review anxiety. We also go a step further to develop and test the effectiveness of a cognitive-behavioral intervention for code review anxiety, so that we can reduce code review anxiety in an evidence-based and empirically-supported way (because science > vibes): https://osf.io/preprints/psyarxiv/8k5a4
What is Code Review Anxiety?
Code review anxiety is characterized by a fear of judgment, criticism, and negative evaluation while giving or receiving code reviews. This is a phenomenon that has been widely acknowledged in industry commentary.
Why should I care about Code Review Anxiety?
One of the primary consequences of code review anxiety is avoidance, which can take the form of ignoring code reviews, procrastinating, "rubber stamping," skimming feedback, etc. This avoidance not only makes anxiety worse in the long-term, but also prevents individual developers, teams, & orgs from accessing the documented benefits of code reviews,
such as improved code quality & security, learning & knowledge transfer, collaborative & creative problem solving, trust, & community building. Additionally, our research shows that, counter to industry myths about code review anxiety being a “junior developer problem,” code review anxiety impacts developers across experience levels (Our sample had 1-65 yrs coding experience). This means that any developer can have code review anxiety, making code review anxiety relevant to any team or org.
What leads to code review anxiety?
Our model shows that we can intervene on code review anxiety by targeting the cost bias (when we overestimate the "cost" of a negative outcome) and anxiety self-efficacy (our belief in our ability to tolerate and manage anxiety), with anxiety self-efficacy being the strongest contributor to code review anxiety.
How can we reduce code review anxiety?
We developed a single-session cognitive-behavioral workshop intervention that was highly effective in reducing code review anxiety, increasing anxiety self-efficacy, and increasing self-compassion. This likely means that our intervention, which focused on teaching cognitive restructuring, primarily worked by increasing folks’ self-efficacy and self-compassion.
@Adora We taught people 1. exercises that help you pay attention to what you're feeling in your body & relax your body; 2. a method to identify & challenge overly negative & biased thoughts you have about yourself and the situation to create a more realistic and kind way of thinking about it/yourself; 3. create a plan for communicating how you feel and asking for/giving specific feedback during the code review
I wonder how much of the result is due to the Hawthorne effect (and related): Not just that the participants knew that they were being studied, but that some of them received the anxiety management intervention, and so expected it to work, and expected that they should manage their anxiety responses.
And also, of course, I wonder about relationships with traditional approaches that are more along the lines of trying to train and convince the code reviewer to not be a jerk about it.
@JeffGrigg@Adora take a look at the methods of the paper! While a placebo effect is possible, it is likely mitigated by the fact that folks in the control condition arrived to receive the anxiety intervention, took outcome measures, and then received the intervention in a single session. If folks were to have answered outcome measures, then received the intervention during a different session, this would be a bigger concern.
@CSLee@JeffGrigg@Adora +++ to what Carol notes, the design of a study like this is a central part of what gives us the confidence (and grounds) to make a claim about efficacy. These types of interventions are also very well supported in other people's well-controlled work in different areas so have the strength of clinical robustness there too. In this case too, people's "expectations" are directly part of the issue in the first place and so modifying them explicitly & knowingly is the goal
I totally deserved negative feedback on that. I did not read the paper carefully.
My lame excuses include that I'm not skilled or experienced in academic papers, and that I was in a hurry to go on a bicycle ride on a nice day before the rain expected to hit mid-afternoon.
Still, I was thinking, while riding, "Good thesis defense, Carol!"
(And that my opinion on that should not be given too much weight, for the reasons I gave above. 🙄 )
We're really interested in finding ways to encourage peer code review in (non-computer science) academic research groups. This paper is incredibly helpful for thinking about how to lower barriers to adoption of peer code review practices!
I shudder to think of all of the learning opps I've passed up because of the code review anxiety I've experienced as a software developer in various contexts. This is the research, workshop and intervention that I've needed.
I just can't emphasize enough how important I think this research is! I hope the #softwaredevelopment community will take the time to read and understand.
@CSLee@grimalkina@seresearchers I found this preprint particularly interesting as a former dev and designer. In the design world, there is a whole culture around critique, including practices that go back a very long way that help mitigate some of the issues outlined here. My computer science education did not prepare me the same way.
@jvschrag@CSLee@seresearchers I thought about such studio culture a lot actually wrt code reviews -- had a lot of exposure to great cultures on this in my design lab postdoc and a much deeper emphasis there on giving good feedback as a learned skill
Add comment