I've been working for about 2 1/4 years on a PhD CS research topic, which is to speed up a certain type of simulation using a certain general approach. This approach is pretty much the only research topic our research group has to offer. However, in my opinion, this approach is also fundamentally broken and beyond repair: It is not applicable or very inefficient for most simulations of this type and for the few simulations, for which this approach is applicable and efficient, there are much better specific approaches available. Consequently, in pratice, nobody uses this general approach over better specific approaches. Besides this fundamental flaw, there are other issues bothering me about this approach and my work, like being pretty much implementation focused, not offering much room for interesting optimizations and that I have to work pretty much alone.

While I have realized all this even before accepting this position, I've been telling my self that all of this will somehow work out. However, as things are now, it's quite the contrary: I've resigned a long time ago and lost pretty much all interest in CS. I've been thinking about quitting this position for a long time, but I do not want to disappoint my supervisor, whom I like very much, and I also do not know where else to work. I'm also afraid of throwing away 2 1/4 years of work on a PhD. What would you advise me to do?

  • What does your advisor think?
    – Bryan Krause
    Dec 28, 2018 at 21:56
  • Well, I think he has at least recognized some of the flaws about our approach but keeps clinging to it, since it's pretty much our only research topic. However, I'm too afraid to talk to him directly about this, since I am very bad at critizing something without hurting somebody. Dec 28, 2018 at 22:46
  • Maybe it's just my perspective, but I have a hard time imagining being in academia without being able to criticize work: that's an integral part of the whole enterprise. I can't quite imagine a group being so wedded to a single approach that they cannot consider other approaches.
    – Bryan Krause
    Dec 28, 2018 at 23:10

1 Answer 1


In general, a proof or demonstration that something does not work for a particular application can be as valuable as proving that something does work. This is especially true if a priori, the thing should seem to work.

You might explore with your advisor whether this is a viable direction for your work to take. I can't predict what he/she would say, but it is worth spending some time in the exploration before you bail out. You probably have some deep insights into the problem that you can bring to bear.

Again, research is about knowledge, not about doing some inappropriate thing just to "force" it to kind of work. Extract the knowledge from the project.

But your advisor is likely the final judge here.

  • Well, its pretty much obvious that if you have some basic knowledge of several simulations targeted by this approach, that it will not work or will not work well. And as you've stated, imho, this approach is pretty much "doing some inappropriate thing just to "force" it to kind of work". Consequently, I find extracting useable knowledge extremely hard: What is the gain of improving the performance of our general approach by x% using some well known technique, if for pretty much every pratically relevant simulation more specific approaches will obviously always be much better? Dec 28, 2018 at 21:48

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .