grimalkina

grimalkina, 8 days ago

Don't judge a problem or solution too early. This is just an absolutely foundational thing: we all struggle to stay "engaged" in working a problem. That does NOT mean "grinding" or being exhausted. It means not being distracted by early errors and being able to overcome them well enough to see a solution. Here, perhaps generative tools could be used as a quick check on certain rabbit holes of error triage. Compare and contrast that with the other methods you have for error triage!

grimalkina, 8 days ago

Elevate the human skills involved in all software work -- this recognition strengthens belonging culture. Being explicit about EFFORT is something many of us don't do enough, particularly if you're in a very "output" driven workplace, making that effort visible might be important advocacy.

Also is a counterbalance: is a tradeoff not being voiced? "this changed how I did x but I saw to make x work we needed y" (like more code...more pressure on colleagues to code review type stuff)

grimalkina, 8 days ago

Honestly -- most software developers are strongly committed to being lifelong learners, but we are STRUGGLING to have learning CULTURES on teams. Tying learning time, effort, and goals to technological innovation is really key I feel. The moment of generative coding tools could be a moment to make this very clear inside of an org. We see learning culture-high devs report higher team effectiveness and productivity. Learning culture isn't a magical thing it comes from deliberate choices

grimalkina, 8 days ago

Let's start with some basics that can help you across MANY situations: misconceptions we have about learning!

Think of using these tools as "active learning," not passive. Passively consuming content over and over again is fundamentally inefficient. We "rob ourselves" of the opportunity to learn in the future every time we don't actively reconstruct. Try generating a part of a solution and then writing the next part yourself, and then seeing if the tool generates something similar.

grimalkina, 8 days ago

Learners with great self-regulation practice reflection. This is another thing we avoid, but where a tiny investment pays off wildly.

Taking fifteen minutes for reflection can wildly strengthen your metacognitive skills. This is how you figure out what your own "deliberate practice" could be. This can be simple. List a small number of things copilot was great at and a small number it wasn't. Explain it to someone else. You'll start to force yourself to be aware of your own mental models

grimalkina, 8 days ago

Because we often FEEL like it's good to "consume a lot of information" (reading a million stackoverflow threads/throwing prompt after prompt into a window), we can waste a ton of time & exhaust ourselves. Kind of like the work version of browsing social media. Instead, choose small targeted effort over hours of meandering and browsing. This is often called "deliberate practice" and planning WHAT and WHY you are practicing a thing is far more effective. Do it sooner than you feel like

grimalkina, 8 days ago

Diagnosing our own knowledge gaps is actually half the battle. Experts aren't people without knowledge gaps -- they're people who excel at recognizing them and solving them early!

I've observed in my research that when exploring an unexpected friction or "break" experts do a lot of "anchoring" -- choosing a piece of the codebase or a causal relationship they feel they deeply understand and iterating in tiny steps out from there. This is in line with good learning science!

grimalkina, 5 days ago

Relatedly? I saw a critique shared in a meme from a fellow complex patient about how surgeons only allow themselves to perceive "current imaging" as your medical history will mean nothing in an OR; I experience this contextless medicine with unevenly weighted labels constantly. If you can imply that you have What Gets on The Plaqueness around your doctors care unlocks like a mechanical animation in a heist movie

grimalkina, 8 days ago

Actually we could be bringing in loads of useful stuff from edtech research into disparities in how new tech usage is even allowed: for instance, the SAME tools rolled out to people with higher power are often used (and people are rewarded for using them) in completely different ways from people with lower power. HOW we are using a tool is a necessary question, and whose usage gets seen as "smart" vs not. The psychological context sets the interpretation of any tool moment.

grimalkina, 8 days ago

But you know, as long as software journals don't even want to allow the reporting of participant demographics, we cannot possibly begin to scratch the surface on any of this. I think it is important to consider the selection effects of what we hear about, the long chains of pipe we've laid down that dictate who is heard in this moment, and what complexity is compressed into legible summaries that treat all "developers" like a neutral blank slate, interchangeable "personas."

grimalkina, 8 days ago

It is not shocking that using Copilot to "creatively iterate" in an agentic way, and for exploratory tasks, is far different from using it in a confirmatory way, for tasks where you lack the skills to evaluate or whether you do test and retest (in a learning sense, not a software sense). The absolute lack of operationalization clarity in this research is really holding us back. WHAT is "using copilot"? We could be drawing from decades of work on how people consume content and active learning

grimalkina, 8 days ago

Why! (jk I know why: those large tech companies that dominate the studies we hear about mostly would not hire anyone with a social science, cognitive science, or real solid learning science background to do software research even if that person was right in front of them because that is seen as less relevant than a CS degree with a vague hobby 'interest' in human behavior, ask me how I know)

grimalkina, 8 days ago

@glyph every single research paper I've submitted to a software journal or conference, the reviewers have pushed back on participant demographics -- collecting them at all, or reporting them. It is neither standard nor behind-the-scenes understood imho. I imagine this might vary quite a lot depending on if you have a pool of reviewers you always get sent to or whatever vs kind of random which is the unlucky position I've been in.

grimalkina, 5 days ago

This is a messy post but with a lot of threads on this: https://www.drcathicks.com/post/expecting-success

grimalkina, 5 days ago

"It’s this environment and the explanations that it encourages that has always fascinated me [...] we certainly have a responsibility to intervene on a bad environment that’s pushing the wrong sets of beliefs.

The thing is, other people matter. We matter to other people. Our explanations of our own success become judgments of other people--hence the runaway success of terribly maladaptive achievement beliefs like "only people born magical geniuses can succeed in STEM." "

grimalkina, 5 days ago

Our beliefs about what it means to be smart. To be good. To have a certain professional identity. What we think communication is, what we think we owe each other (or not). Whether or not we believe other people will see it like we do and what we think will happen if they don't.

grimalkina, 8 days ago

But guess what -- you can donate to this program if you wish!!

You can learn more about the program overall here: https://ugresearch.ucsd.edu/programs/all-urh-programs/start-neuro/index.html

They have a fund through UCSD: https://giveto.ucsd.edu/giving/home/gift-referral/8dfa8984-7da2-4d59-b762-92621fa6228e

I'm not involved in any way (except for sometimes hosting the scholars at a bbq at our house haha...and a lot of support to the program director), I just think this program is absolutely beautiful and more people should know about it. The future of science is for everyone.

@analog_ashley