The MANDA Lab Odyssey

A few quick observations.  When talking to MS or PhD students during seminar presentations or other scientific meetings, I sometimes notice that they answer questions about very concrete topics (e.g., mathematical models of memory) quite confidently and fluently .  However, fluency is primarily verbal in nature. A lot of fluent sequences of words but without a genuine sense of understanding. I wonder if that is partly an LLM effect. All of us use AI these days. When you ask AI to explain something, it does so fluently but not always in the simplest words.  The only solution to this that I can think of is the old-school pen-and-paper approach to reading papers, writing down equations and struggling to make sense of them, drawing simple pictures of how you think matrices transform through the equations of the models, etc. This is, of course, time-consuming. An LLM will spit out the answers in seconds, and you read the answer and go away with the feeling that you've understo...

 I was asked this question recently in a podcast done by the institute. This is a podcast that replaces "reflection talks" that faculty used to give before their 5y contract renewals or promotions.  The reflections podcast, with Richa Chadda, senior manager at the faculty affairs office. I used to accept people to the lab simply on the basis of an interview to gauge interest. I thought the rest could be discerned from their CV. Unfortunately, this quickly backfired because CVs don't reflect reality. More often than not, CVs are exaggerated. In other cases, they are filled with internship experiences but each lasting a month or two. In other words, there is a distinct lack of depth and commitment shown in most CVs but people are quite articulate these days and a simple interview is insufficient to gauge preparedness to take on research in cognitive science. The other issue is people think research in cognitive science is easier than say in quantum physics. So many people w...

Universities and teachers around the world are being forced to rethink their approach to teaching given the rapid advancement of AI. Most homework and assignments can be done by AI far better than even the instructors of the courses. AI-detection tools are not the answer since they will always be gamed, and moreover, preventing people from using AI isn't really training them for the real world because AI is here to stay. We have AI on our phones as pocket companions. Google searches have been replaced with AI answers to queries.  Viva voce to test students instead of take-home assignments, handwritten term papers in class, etc. have all been suggested, and I've tried them myself. However, these are interventions at the point of the test and not something that really encourages students to value learning. AI-driven grade inflation  does not imply better learning; in fact, the opposite is most likely true .  I happened to come across this short but interesting piece on Medi...

I will keep this short.  The sign of a great leader is their willingness and ability to lend wings to junior members of their team so that they can fly and achieve their potential.  Unfortunately, what I see all around me are overwhelmed people in leadership positions who keep resources and opportunities to themselves despite their inability to take more on and be able to do a good job with any of it. This behaviour may not even be intentional. This is what they probably had to do to survive in a low-resource environment. However, when everyone does this, we collectively stifle scientific progress by not allowing meritorious young researchers to do the best possible work. Academia is not meant to be a dog-eat-dog world, but that is what it has turned into.  I do see rare people who are true leaders in academia. Without exception, they all give support and resources to people who are junior to themselves and pass on opportunities, especially when they themselves are not fu...

I have had many frustrating experiences with mentees and others who are smart but overrely on AI for their work these days. The reliance on AI for coding is not the issue. The problem is that they set their human brains (which are typically quite adept!) aside when they work with AI, and over time, this sets in real rot of critical and deep thinking. When candidates interview confidently and answer hard interview questions, a second round comprising a data analysis challenge usually reveals the issues. During this second interview, I ask conceptual questions. Only if the applicant has taken some minimal effort to understand the code and the concepts involved in the analysis and has thought a bit about the cognitive questions driving the data analysis challenge can they pass the bar for entry to the lab. Ten years ago, most people with master's degrees (and sometimes research experience in the same domain!) from IITs and other reputed colleges and universities would have been able t...

On June 30, 2026, I will complete 5 years of running a human memory research lab. Here is a summary of our research efforts so far and vision for the upcoming 5 years. Overview of our research program Our main area of focus is human memory. However, we are investigating memory at multiple levels: 1) Memory at the stimulus level (averaged across many people)— memorability 2) Memory at the neurocognitive mechanistic level— temporal context and event segmentation 3) Neural propagating waves as the basis for multiscale neuronal communication (and therefore, cognition)   in the brain. Figure 1. An overview of our program (thanks to ChatGPT) for those without the patience to read the text (written without ChatGPT). :)  1) Memorability We have developed a body of work describing what makes experiences memorable. For instance, the word " PINEAPPLE " may be more memorable to people in general, regardless of their individual memory capacities, than a word like " FOOT . "  ...