1

I just started advising graduate level students, I am having a lot of problems. I thought I would be a patient, understanding, helpful advisor but really I am being tested. One of the students only copy things, she did some literature review and presented it, fine, but when I ask the finer details it is clear that she hasn't grasped concepts that she presented. I talk about the problem that we will work on and hope she will grasp the big picture, but then when she struggled I thought maybe we should start small. So I gave her a machine learning task, straightforward task, gave her resources. Nothing. She says she doesn't understand how it is related (literally only difference is the dimensions of the data, but she has difficulty grasping high dimensions of data and doesn't see it is just a matter of array size). Then I found her a tutorial, step by step, which uses an identical format of data, I asked her to try to understand and write a report so we can talk about it (when I don't ask for something concrete she just doesn't do anything, I asked her earlier for progress email exchanges and she just went missing). For this tutorial, where each line of code is explained, each line of data is explained, she still asked a million questions (only on the format of the data, nothing about machine learning part). I answered patiently, and asked her to please please try to "understand" what we are trying to solve. About her data questions, I sent her a million links to show all she has to do is "google", it is a famous dataset, there are even discussions about it on kaggle. Honestly, this blows my mind because this is not something even undergraduates should ask. Just file operations, she asks "do I use this file" or that file as input without considering what we are solving. So for the report, she sent it the next day, seemingly just translated the tutorial. And I explained her, it is not homework, but it is to help her understand. She said she can't meet online when I offered to talk about what she learned etc. Anyway, I asked her to move on to our dataset keeping this example in mind. And she asked me a million data format, file reading questions. For the sake of moving on, I wrote a very detailed guide (which I really shouldn't have to), and now she did like one step and asks me if it is correct. I followed a similar route with another student, and when we did a meeting on a small task he was supposed to work on (image classification), it was clear that he learnt -nothing- at all although he had running code. I mean nothing about how the example works. He just repeatedly talked nonsense about input/output and neurons mimicking the brain when I asked what a convolutional layer does. I said it is okay to say that he doesn't know if he doesn't know, and I explained it to him. But, all book and video resources I sent him has this information and he had 2 months on this. The first student was assigned to me because I was told she had a high GPA and I needed to graduate a graduate level student asap so that I can rise etc etc. (Really not something I care about, but the school does this for everyone who just started a position) She seemed very eager but I am beginning to doubt she has ever written code for a project, on the other hand I accepted the other student because he was so keen to learn machine learning, and also because he is an international student( not many people in the dept are open to international students other than me). None of these are project hires. I try my best to tell them that they need to do research, they need to work on novelty, they need to understand the theory and inner workings of machine learning.

**- Can I outright tell them that there is a limit to what I will answer? That I am not expected to answer all their questions? And I really shouldn't? What is the opinion on this? I don't think for a graduate level student, I should be wasting time on file operation. More importantly, I need the students to think on what we are trying to solve, so that they will know what they need as input. I want them to think about it and understand it, I showed them the way but they just won't walk from there. How to not be a jerk but communicate this if they aren't responding to my requests for them to try to understand what we are trying to solve, draw connections with other works, study inner-workings of theoretical concepts **

  • Obviously, I can't give a whole lecture on the topic, so they need to learn these on their own.

  • My plans for a journal club seems delusional right now.

On another note, I have undergrads who had failed the graduation design project before (so not the typical "best student" profile), but passed the last semester with flying colors when they worked with me, and they never tired me as much. We worked closely and they were able to complete all the tasks, understand inner workings, even add their own contributions. I made them explain each line of code they wrote, when I asked them to make changes in a short time they were responsive, and they thanked me for my interest and for helping them learn from scratch.

  • 4
    This sounds like a terrible situation to be in, and I'm not sure what anyone here can say that would be of much help beyond providing some sympathy. However, there are some things you could include to provide a bit more context. Are you new to this university? If so, what was it like where you were at previously (whether as a student, postdoc, or teacher)? What have your colleagues said about these concerns? And if you haven't asked any, why not? Are these Masters students or Ph.D. students? Do your students have qualifying exams they need to pass? If so, then maybe that's their main concern. – Dave L Renfro Sep 19 at 12:51
  • 1
    Yes, I am new to the university. I did my MS and PhD in US, in an ambitious lab so it was quite different. I am aware of the limitations and I don't expect the same from them as my supervised expected from me. I literally started the position with covid, so my exchanges with the colleagues were limited. One of them tells me that our students are not ambitious so I should lower expectations, another only accepts students that he had lectured (which makes sense now). These are MS students, so I am aware they might be academically inclined. – dusa Sep 19 at 14:18
  • Lowering expectations is okay, but there is no way we can progress if we can't move past a few file operations needed to prepare input to a ML model. I honestly don't know how she has a high GPA, my guess is that she was just lucky in group projects, because if anything, she needs to complete a graduate design project for the undergrad (and possibly a lot more in courses). And she said she did a ML project last semester, I kind of checked what she did and saw all code was pushed to github by another project member. – dusa Sep 19 at 14:19
  • There is no qualifying exam for MS. She presented her literature review and it was enough for the department. Just after a week of presentation she didn't know the details she presented. Not all students are like this btw, another student had already started working on his method and showed us the demo of what he had so far. I have to put some ground rules I suppose on whom I will advise, but students need to find advisors eventually. – dusa Sep 19 at 14:23
  • 1
    I actually think the journal club might help. Regular meetings prevent the disappearing act. Another trick is to just be less accessible. Get the students used to the idea that you'll help them X number of times between meetings, and expect results at meetings (X can go to zero when they get stronger. I'd start at X=2). – Well... Sep 19 at 17:42
5

Background (okay to skip): I myself am a PhD student and an undergrad recently started working with my advisor on a senior thesis project. My advisor is pretty much absent, and works by throwing lots of papers at someone and then leaving them to their own devices. Most of the time she can't answer my questions so I don't even try. Not because she's incompetent but because she doesn't really even communicate much with me so what I'm working on has only a little bit to do with what she knows about. I decided to meet with the undergrad once a week to attempt to provide the guidance I wish I'd had when I was in his shoes and he matches the description in your question.

I think there are two main problems:

  1. Many undergrads who have stellar grades have learned to play the grades game without learning very much critical thinking. My undergrad GPA pales in comparison to my new "mentee" but I am positive I am better at googling things I don't know and persevering on problems that are way above my head.
  2. You might be surprised by how little the students actually understand. I constantly felt overestimated when talking to older mathematicians. Most of the time, during meetings, conferences, and even classes, I had absolutely no idea what people were even saying. You also can't assume a student knows the material in classes they took. They might just be in it for the grades. I think my mentee makes very little progress and asks too many questions simply because he has been given papers that are way above his head and he's not ready for them yet.

To make a long story short, your students seem to lack bravery. They need to build confidence and you can't make them do that.

My idea of a solution:

  1. Set your expectations straight. Giving clear guidelines for questions might help but I don't think it's enough. I think if I were your student, I would appreciate hearing what you've written in this question. Tell them what you think of their progress. Don't be mean, but spin it like this: you students have a lot of potential and you're really intelligent but you're just not going to learn anything or make any progress if you keep coming to me every time you're stuck! Also make it clear that PhD students are kind of like assistants. You can guide them with how to learn but you aren't their lecturer. They have to learn on their own and they should come to meetings with something to contribute.

  2. This whole "disappearing" thing can't be tolerated. You're obviously putting a lot of work into mentoring them. Make it clear also that they need to give you reports even when they don't have a concrete goal that you handed them. I would love it if my advisor asked for that! But she doesn't and I have to put all the effort into the relationship. But maybe the biggest part of research is finding a good question to ask and a good problem to work on. They also need to learn to do that. Giving you a weekly status report will help them keep themselves on track, reflect on their progress, and give you an idea of what they are doing but it should not be a lot of effort on your part.

  3. I would give them a small task or problem that you think they should be able to solve themselves and tell them you won't advise them until they solve it, without help. I know a professor who doesn't take students until they can solve any question in a particular (difficult) textbook on the spot. You don't have to be that extreme, but just give them a problem you think is instructive and you know they should be able to handle at least eventually. Assuming this is completed successfully it will go a long way in building their confidence.

  4. I would also make it clear that it's okay to take some time to learn the basics, something my advisor never did. But it seems like you're already doing that.

If they students respond badly to this, honestly I think you've done everything you can. But at least this gives them a chance.

| improve this answer | |
  • 1
    I relate, as in, sometimes I did give the impression that I knew things when I didn't have a clear understanding. And at times people were surprised I did better than the impression I gave. I had a pretty isolated PhD as well where I was left to my own devices. I was in an ambitious lab but there was not much guidance for the most part, particularly for my project and it wasn't a time when internet was booming with frameworks, tutorials, implementations. – dusa Sep 19 at 14:34
  • 1
    So, exactly I wanted to be a better advisor, and I am trying with long discussions even on weekends, but now I am at my wit's end. It is what it is. Like you said though, it is not the knowledge per se, but rather a trait. Thanks for the tips, I did most of them the way you suggested and don't see it working but at least I feel reassured, so thanks for sharing them. And I definitely need some rules on whom I will advise I suppose (#3) – dusa Sep 19 at 14:35
  • To add to my earlier point, I can see how setting some rules or giving tasks before I work with a student can solve this problem, but I am not sure if I am comfortable with this. I was thinking having an interest should be enough. Then again, it is better for all parties involved to figure out the nature of study early. – dusa Sep 19 at 17:37

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.