Item — HN Reader

> Using ChatGPT on cognitive tasks can reduce your brain connectivity by up to 50% and reduce your ability to recall information about the task by 8x.

Argh people keep referencing this study as Gospel. It has not been peer-reviewed. Its methodology has a number of concerning confounders. It's a tiny sample with a narrow contrived task domain. And the very premise of the study is misframed. The implication that 'brain activity' is a positive outcome does not follow. Brain connectivity might be analagous to inefficiency as opposed to the reported 'engagement' or 'cognitive debt'.

7 comments

I agree that it's not a great study, but I also don't want to find out too late that it wasn't a good idea to outsource my thinking to ChatGPT.

The study wasn't great but don't outsource your thinking.

The study (if we assume it was good) told students the objective of their task was to generate a factual essay about topic X. The study then measured how much they learned about the topic at the end, but the students who used ChatGPT "learned" less and remembered their own essay less despite an equally passable essay. I want the alternative version of the study where the students are told the essay is practice, and their success will be graded on how much they've learned about the topic.

I imagine you could conduct a similar study challenging students to complete math tasks with and without a calculator and then ask them how much of their multiplication tables they've learned afterwards.

If you want to learn and grow as a person, along some dimension, you need to practice. Growing requires repetition and reflection and to experience the feedback loop of improvement. Outsource thinking for whatever task you don't want to do when the only result you care about is the only outcome. Don't outsource practice and learning if you want to improve. Only you can make the decision on when each situation applies in your day. Maybe you want to be better at some task at your job, but maybe you just need to get through the task and move on.

Speaking about not outsourcing practice reminds me of the physical analogue...

I have heard about people talk about "farmer's strength" to reference a very natural functional strength that is earned by the gruelling and diverse physical demands of doing farm labour for a lifetime.

Now people have invented various training regimes to try to reproduce that kind of strength outside of the original farming environment.

Edit - As an aside, it just occurred to me that I am both a functional strength and functional programming proponent (facepalm). Perhaps in the future after people seeking to strengthen their minds through via mental gymnastics, FP will see a renaissance

It's kind of the opposite of strength training regimens, it's more like people saying "Why would I pick up the weight by hand when forklifts exist? When, in the real world, when doing a job, am I going to be expected to pick up dumbbells by hand lol? I'll just use a forklift every time."

Yeah, this isn’t the first time we’ve outsourced our thinking. The ubiquitous use of GPS comes to mind leading to lower abilities in the hippocampus.

It grows when a person trains to be a London cab driver (no GPS allowed). [0]

Yet abilities decline as GPS use is used. [1]

These aren’t huge studies either, but also fall under the larger, “use it or lose it,” umbrella.

[0] https://pubmed.ncbi.nlm.nih.gov/34914151/

[1] https://www.nature.com/articles/s41598-020-62877-0

I find it completely acceptable to outsource some of my thinking and tasks to LLMs. Not everything is my primary area of focus or expertise, and not all questions are equally important. For example, I could buy any random toaster oven, or I can kick off a deep research to pick a best toaster by my criteria and in 5 minutes I’ll have a much better context for this decision, without having to switch my mind from my primary objectives. I don’t want to think about toasters.

So what, ChatGPT is the new weed? One critical statement draws out all the addicts in defense?

ChatGPT is the new weed: everyone has a source they cite to reaffirm their pre-existing position.

Sure, but “computer does all of the logical inference by hand and it turns out your brain isn’t needed” is awfully a big confirmation bias in favor of people who have left LLMs work in their favor.

Same for the reference to developers taking longer to complete tasks with AI. That's absolutely not the finding of that study.

Im never surprised when someone prattling on about dopamine also leans on a bad study to make a spurious point.

Yep it's always a bit cute and funny, when you consider the absolute necessity of dopamine in basically every functionally relevant neural activation. Talk to a parkinson's patient about your 'dopamine=bad' fluff. Ugh. They may as well just have titled it "Protecting My Attention At The {Insert Arbitrary Hormone or Neurotransmission Chemical} Carnival"....

> They may as well just have titled it "Protecting My Attention At The {Insert Arbitrary Hormone or Neurotransmission Chemical} Carnival"....

I disagree. I am absolutely certain that the vast majority of the readers here would have known in the context of that headline exactly what "The Dopamine Carnival" meant, without needing any specific positive or negative implications about dopamine in general or it's actual biochemical mechanisms. It's blatantly obviously about social media and mobile apps that are intentionally designed to manipulate your brain and its reward system.

Ofc the title got its point across, but I'd argue to hold ourselves to higher standards of veracity. That's all.

I am sure studies would find I use my hamstrings and quadriceps less when driving than when walking too.

The problem is if you are driving a car to go the same speed and distance as walking then it is missing the point.

Looks like another ChatGPT victim here folks.