nataliyakosmyna.bsky.social
Research scientist @mit @fluidinterfaces @mitmedialab
Ph.D in Computer Science and Brain Computer
Interfaces.
Project lead @NeuraFutures, Augmenting Brains
23 posts
107 followers
17 following
Active Commenter
comment in response to
post
Unfortunately press ran with it on “brain damage”, “brain scans”, “harm”, “dumb”, which we never state…
comment in response to
post
Thank you for your feedback! Points well taken, as the sample size is small! We will be able to add some of the details you are pointing to, but for some - we would need to reiterate what we are saying already: limited nb of subjects + specific task.
comment in response to
post
Also, shoutout to these authors for including a section on energy usage.
comment in response to
post
Hello! Which tools do you have in mind? Our logic was to compare it to the closest you would use before an LLM, for a similar task, which is to do a web search. Also, a question about quoting was just one of the questions.
comment in response to
post
I’m refusing to on board to Copilot at work. I’m the only person. When asked why I sent them the MIT and Microsoft research papers as openers.
comment in response to
post
Oh no, thank you, no AI use. 😬
comment in response to
post
:-)
comment in response to
post
Huge thank you to my amazing co-authors: @eugenehp.bsky.social Tina Yuan Jessica Situ Xian-Hao (Harry) Liao Ashly Vivian Iris Braunstein and Pattie Maes
comment in response to
post
𝐈𝐕. 𝐁𝐞𝐡𝐚𝐯𝐢𝐨𝐫𝐚𝐥 𝐚𝐧𝐝 𝐂𝐨𝐠𝐧𝐢𝐭𝐢𝐯𝐞 𝐄𝐧𝐠𝐚𝐠𝐞𝐦𝐞𝐧𝐭 (part 3)
Cognitive Debt: Repeated LLM use led to shallow content repetition and reduced critical engagement. This suggests a buildup of "cognitive debt", deferring mental effort at the cost of long-term cognitive depth.
comment in response to
post
𝐈𝐕. 𝐁𝐞𝐡𝐚𝐯𝐢𝐨𝐫𝐚𝐥 𝐚𝐧𝐝 𝐂𝐨𝐠𝐧𝐢𝐭𝐢𝐯𝐞 𝐄𝐧𝐠𝐚𝐠𝐞𝐦𝐞𝐧𝐭 (part 2)
Critical Thinking: Brain-only users cared more about 𝘸𝘩𝘢𝘵 and 𝘸𝘩𝘺 they wrote; LLM users focused on 𝘩𝘰𝘸.
comment in response to
post
𝐈𝐕. 𝐁𝐞𝐡𝐚𝐯𝐢𝐨𝐫𝐚𝐥 𝐚𝐧𝐝 𝐂𝐨𝐠𝐧𝐢𝐭𝐢𝐯𝐞 𝐄𝐧𝐠𝐚𝐠𝐞𝐦𝐞𝐧𝐭 (part 1)
Quoting Ability: LLM users failed to quote accurately, while Brain-only group showed robust recall and quoting skills.
Ownership: Brain-only group claimed full ownership of their work; LLM users expressed either no ownership or partial ownership.
comment in response to
post
LLM-to-Brain participants had limitations in achieving robust neural synchronization essential for complex cognitive tasks.
Results for Brain-to-LLM participants suggest that timing of AI tool introduction following initial self-driven effort may enhance engagement and neural integration.
comment in response to
post
𝑺𝒆𝒔𝒔𝒊𝒐𝒏 4 𝑹𝒆𝒔𝒖𝒍𝒕𝒔
LLM-to-Brain (🤖🤖🤖🧠) participants underperformed cognitively with reduced alpha/beta activity and poor content recall.
Brain-to-LLM (🧠🧠🧠🤖) participants showed strong re-engagement, better memory recall, and efficient tool use.
comment in response to
post
𝐈𝐈𝐈: 𝐄𝐄𝐆 𝐀𝐧𝐚𝐥𝐲𝐬𝐢𝐬
Connectivity: Brain-only group showed the highest neural connectivity, esp in alpha, theta, and delta bands. LLM users had the weakest connectivity, up to 55% lower in low-freq. Search Engine group showed high visual cortex engagement, aligned with web-based information gathering.
comment in response to
post
𝐈𝐈. 𝐄𝐬𝐬𝐚𝐲 𝐒𝐜𝐨𝐫𝐢𝐧𝐠 (𝐓𝐞𝐚𝐜𝐡𝐞𝐫𝐬 𝐯𝐬. 𝐀𝐈 𝐉𝐮𝐝𝐠𝐞)
Teachers detected patterns typical of AI-generated content and scoring LLM essays lower for originality and structure.
AI Judge gave consistently higher scores to LLM essays, missing human-recognized stylistic traits
comment in response to
post
𝐈. 𝐍𝐋𝐏 𝐚𝐧𝐝 𝐄𝐬𝐬𝐚𝐲 𝐂𝐨𝐧𝐭𝐞𝐧𝐭
LLM Group: Essays highly homogeneous within each topic, little variation. Students often relied on same expressions or ideas.
Brain-only Group: Diverse and varied approaches across students and topics.
Search Engine Group: Essays shaped by search engine-optimized content
comment in response to
post
For 4 months, 54 students were divided into three groups: ChatGPT, Google -ai, and Brain-only. Across 3 sessions, each wrote essays on SAT prompts. In an optional 4th session, participants switched: LLM users used no tools (LLM-to-Brain), and Brain-only group used ChatGPT (Brain-to-LLM).
👇
comment in response to
post
Thank you for explicitly mentioning the Search Engine group, most people only actively discuss LLM vs Brain-only groups!
comment in response to
post
I appreciate you reading it! It is a long read! Thank you!!
comment in response to
post
It is!
comment in response to
post
If you are a human - please do read the TL;DR we added!
comment in response to
post
We included a TL;DR and a summary for your favourite LLM (try to use TL;DR though, we recommend!)
comment in response to
post
:-)
comment in response to
post
Hahah, ok, please do not hate me, but I am not a fan of Latex! I have so many issues with it from UX/UI perspective (I’m the lead author on the paper).
But tell me - what is giving it out?
comment in response to
post
Submitting! Cutting it as we speak so the SoA is coherent between multiple submissions as unfortunately, this paper draft is too long for a single submission :-/ We did feel the need to release it all as one document though, as the results are coherent and brain/NLP analysis do correlate.