The Independent Science Society Board

https://essays.papertlab.org/p/thoughts-june-2026

Papert Lab is an applied cognitive science research nonprofit that uses experiments to model how humans solve problems and explores the implications for designing computational mediums that augment memory, learning, and curiosity. We do this through designing intelligent tutoring systems and exploring new interaction paradigms between artificial intelligence systems and humans.

Our current research focuses on characterizing how novices design specifications when interacting with agentic coding systems. We ground these empirical findings in previous theories of scientific discovery and productive struggle. We submit to HCI and computing education venues, run qualitative think-aloud studies, pilot new pedagogical paradigms in undergraduate courses, and host weekly discussion forums on augmenting cognition with computers.

Papert Lab was founded in 2025 by Hudson Mitchell-Pullman, an independent researcher and high school student. Hudson collaborates with and is mentored by David H. Smith, Annapurna Vadaparty, Qianou Ma, among others.

I enjoyed reading the Papert Lab introductory essay. In particular, I had never heard of the phrase constructionism, and I'm trying to figure out what that actually means. I had also never heard of the term specification design, and I think the specification design questions raised in this essay are worth digging into. To me, I took specification design to mean the skills needed to interact with AI coding agents properly, in a way that one's cognition is not eroded. This means means somehow figuring out the right questions to ask it so that you are learning something when you build something. More details below.

---
In the first sentence of the essay, Hudson writes that "Seymour Papert developed constructionism as a theory of learning to support his vision that computers would permeate the lives of all people."

It was unclear to me exactly what constructionism is, because I've never heard of the term before. I would have appreciated a bit more context into how constructionism is defined. It seems that constructionism was a theory of learning and education between 2000 and 2010, based on the idea that people learn by actively constructing new knowledge rather than learning existing knowledge. So, for example, maybe I would have learned the theories of gravity better by trying to derive them myself and writing an essay about it, instead of learning about Newton's laws.

So, therefore, unfortunately, it's quite hard for me to have a reference point for what constructionism is, because I haven't built a mental model of exactly what the lab means by constructionism. But from here on now, I'll take it on faith that it roughly means to learn something through building versus learning by remembering.

Bear in mind, though, that I'm not an expert on the history or theories of learning. So it could be that this essay is meant for people who are not beginners to this kind of thing.

In the second paragraph, though, I do agree with the arguments that computer science courses in high school are boring, and that it's likely the overly formal approach to learning computer programming — through syntax and scattered concepts — is not inspiring enough, and misses the wood for the trees.

In the "Is accessibility always positive?" section, I really like the idea of the first paragraph, where the lab aims to make humans feel empowered and excited to use things to augment their cognition. I've been trying to think hard about where AI is meant to lie in the scale of how I learn things, but I'm particularly excited about philosophies focus on human enhancement rather than human replacement. Emotionally, I feel bad that computers are smarter than I am in many fields. However, I'm excited that I can learn things quicker. One risk, of course, that I've been thinking about with using an AI to enhance my cognition, is that it can easily and accidentally slip into replacement. I'm not sure how this might happen, and what I'm allowed to delegate to an AI, but it's been on my mind for quite a while.

Because of this, I sympathize with the concern that accessibility could be bad. In the third paragraph of the "Is accessibility always positive?" section, Hudson writes that it is problematic that non-technical people can still use computers even though they are not cognizant that they lack technical skills and are at a disadvantage. So, yes, I agree with the issue of skill deficiency. And therefore I agree that GUIs do inhibit constructionist learning (in the way that I understand what constructionism is).

On the "Specification design" section, I thought it was a bit unclear what the essay is saying. I could have used a more clear example of what specification design is.

Right now, the way I understand it, specification design would be a suite of skills that a student could actually engage with an AI agent for. So, for example, I think this would be somehow giving the student the necessary mindset to know what the right questions to ask about the program are, to make sure it's robust. Or it would be figuring out how to get the student to ask the right questions to the AI, to make sure that they actually know what the AI is making.

On the **Specification Design** part

So here are some examples of what I think specification design skills are:

- The student knows to ask the AI when these programs could fail.
- The student knows the right questions to ask the AI such that they know the structure of the program and how it works to first order.
- The student knows to ask the AI what the risks are of using such a program.

On the **How Undergraduates Probe Scenarios** part

I think it was a bit unclear what this sentence was saying: "*One critical aspect of specification design is the ability to recognize the objects and their attributes embedded within a scenario, and to use these structural components to iteratively design and articulate the intended behavior of the system."*

I do think, however, that the experiment where they got students to think out loud whilst doing the interviews was very interesting. And I'm really excited to see the in-house tools that they used to conduct this study. In particular, I'm quite interested to see what questions these undergraduates are using to figure out if their programs are correct.

Had a cool discussion with Hudson about this on discord, looking forward to getting a clearer understanding of what spec design is in a future essay.

Love this! I wonder if you could use devices that add to the signal of what you're learning about to integrate it more deeply. For example, if you're learning about waves, and solving an equation for a wave that is a specific shape, would wearing a haptic belt that is "playing" the vibration of that wave help you better understand how different waves but with similar variables behave? Or even simpler: if you're doing multiplication and division problems with lots of different types of numbers (whole numbers, negative numbers, very tiny numbers with a decimal point and lots of zeros) and you're wearing the haptic belt, as you focus on each problem, the belt "plays" an ascending or descending vibration, so that you at least get a _feel_ for and learn to associate the shape of the product of some calculation before even solving it. Perhaps this would be helpful for better understanding the patterns of the world/knowledge/everything, and to make faster inferences.

Haptics are just one example. You could do this with many different senses. But I like haptics because you can represent things spatially (you can do that with sound too, would be fun to play around with.)