Posted on July 4, 2019 by Gabe Parmer
In an ideal world, each PhD researcher1 would enter into a department and into a relationship with their advisor that is centered on their continual development as researchers. This is not always the case. Like any relationship, the PhD researcher/advisor relationship can go very wrong. Often, it is hard to know when it has gone wrong, and what to do. I’d like to give my perspective on what students should rightfully expect from a PhD environment – a bill or rights of sorts – and on some thoughts on what to do if a student finds themselves in a toxic situation.
I am not a perfect advisor, and I have made many mistakes. Thus, this is only my opinion and perspective. I believe it is useful as this is a market with extreme information asymmetry, so I figure that having more information and options accessible is a good thing.
This is a very long post (~6000 words), so feel free to read the sections that are most interesting to you.
The goal of a PhD is to become a world-expert in a topic, create new knowledge in that area, and learn how to effectively communicate and disseminate that knowledge. PhDs often take between 4 and 8 years (depending an many factors) because it involves
Each of these activities are very personal. Each individual will approach them differently, and will most productively learn with specialized coaching. This is part of the reason that PhDs are based on the mentor system: an advisor leads the rising researcher through the salient work in a field, how to dive into new research problems, and advises them on how to present their work in a specific community. This is a relationship that serves as an effective means to enable the creation of world-class researchers.
Unfortunately, this relationship can go wrong. Most of academia is based on the assumption that professors are doing the right thing for students and advisees. Most of the time, the system works. However, there are core academic incentives that can negate this assumption, and there are bad people in every field that might take advantage of their students. When a PhD student finds themselves in a bad situation, it can be exceedingly difficult to know what to do, or to feel “boxed in” with no power. It is important to understand what to expect from an advisor/researcher relationship, if for no other reason than to know that the situation you’re in is often not your fault.
What should every PhD research expect from their department, and their advisor? As a PhD student looking to complete their doctorate, I believe you should expect at least the following.
Respect. To be treated as individuals with the respect of being on the hard path to becoming first-class researchers. As you learn about the depths of a domain, there is a phase where everyone starts to feel that they know very little about CS and about research. This is quite intimidating and can challenge your confidence. This can lead to slower progress than you may like as you wrestle with complexities and nuances. When you write papers, start designing systems, or start running experiments, it is hard to remain scientific (though this may help) and have confidence in your direction and decisions. This is all natural and part of learning new, challenging skills.
Your advisor is there to lead you though this process, point out where things can be done in a different or better way, and explain the reasoning as to why this is the case so that you can learn and become increasingly capable. If you feel like your advisor is berating you, making feel fundamentally incapable or insufficient, or generally degrading you sense of self and your mental well-being, this is not part of the normal path of learning. They should be there to help you improve your capabilities, not shame you if there still are improvements to be made.Advisor as mentor, not boss. A research advisor is not a boss. A boss can tell you what to do within the broad parameters of your job, and what the company needs. A research advisor is a mentor - someone who should walk you through the skills, processes, and methodologies necessary to successfully conduct research in the chosen domain. If you don’t want to work on a project, you don’t have to. If you don’t want to make a deadline to protect your sanity, you don’t have to. If you don’t want to do work in a manner that fails meets your ethics or standards, you don’t have to. Advisors cannot tell you what to do, full stop. They can mentor you, collaborate with you, and convince you of the best direction to take.
There are some trade-offs, of course. If you don’t want to take part in the commitments that an advisor has made (i.e. work on funded proposals), or want to work on topics outside of their interests, they have the option to not commit their time and resources to you. Our grants fund our research in specific areas, and we cannot pay for research outside of those areas. This might simply be a situation where you and your advisor are not a good match. This is not uncommon. In some sense, it is amazing that researchers can come together with similar interests as often as we do! Please see the section below on parting ways with your advisor.Choice and a threat-free environment. I know of cases where a professor coerced a student to do a task that they did not want to do, under the threat that their VISA will be terminated and they will be deported. This is obviously wrong, and is one of the best arguments I’ve heard that tenure is broken. I’ve heard of cases where the Doctorate degree was held as a means to coerce students into doing tasks that didn’t relate to their thesis. Threats – including those that are this obvious, and those that are subtle – have no place in academia. Threats are different from statements of fact that could take the form: “if we don’t make this deadline, we might get scooped”, or “if this result is correct, we need to re-evaluate the approach”. Facts should be taken at face value, and often propose a decision point (“do we all want to push for this deadline?”), or require additional investigation (“lets push into the results, and be willing to accept that our hypothesis was wrong”). Threats, on the other hand, often come with a lack of choice. If you feel like you’re put into a situation where you must do something, even if you don’t want to, then the problem is the situation and how you were put there. If you feel like you’re forced to fabricate results, or represent them in a misleading way, the problem is that you’re being forced to do that. If you feel like you’re forced to sink unreasonable amounts of time into work, the problem is that you’re being forced to do so.
You should always feel that you have a choice, and that you’re free to make it without negative, malicious repercussions from your advisor. Your advisor should be there is there to support you, and sometimes it just isn’t reasonable make a push or get some results. The discussion should be about how to proceed after the decision, not about why you shouldn’t have the choice.Collaboration specifications. Collaboration on research should be respectful. You should feel confident that if you present an idea or a direction that it will be considered and seriously discussed. If it is not chosen as the direction of the research, you should know why, and see the reasoning behind the decision. If you feel like you are not being heard, or are only there to listen to your advisor tell you what to do, that is not a collaborative, mutually respectful environment. See my discussion of psychological safety for more detail.
All researchers (including senior researchers) must collaborate appropriately with other researchers on a project. When you publish a paper, there is often a lead author (or small set of lead authors) that drive the research, a number of supporting authors that contribute to the research direction often through comparison cases, applying the research in a number of domains, and, generally, running additional experiments, and the senior researchers that at least mentored the process to completion. Different communities, and different research groups define these sets of researchers differently.
All researchers in a collaboration should receive a benefit from that collaboration, and should have to contribute to it. See the ACM authorship guidelines for an example of how to think about this. This breaks in a few cases:Research group responsibilities. It is typical for PhD researchers to be expected to integrate into and contribute to a research group. This takes many different forms in many different groups. This can include helping run research discussion groups, helping more junior researchers, and providing assistance to other researchers in infrastructures that you know well. You should have a discussion with your advisor early on as to what these responsibilities are, and if you don’t want to take part in them that should be a discussion. The result of not wanting to take on some of these responsibilities might be that the research group is not a good fit for you. This is why you should have this discussion early on, so that you can pursue other opportunities.
As an example of some of these responsibilities, in my group we try to focus on civil, productive discourse, and we’re all (me included) required to contribute to our shared software infrastructure. This means putting time into ensuring that our research results are shared in the infrastructure. The side effect of this is that each researcher is able to stand on the shoulders and contributions of past researchers. But it also means that time must be spent maturing the research implementations, and contributing them back to the infrastructure. If you’re in a group that really builds systems, this is a common requirement, but has the benefit that everyone in the group shares the contributions of that system.Expectations and parting ways. A PhD researcher should understand what is expected of them in terms of research, skill development, and research group involvement. Where the expectations aren’t clear, you should feel welcome to ask for clarifications or for a meeting to plan your goals. If you don’t meet expectations, you should still expect a supportive environment and approach from your advisor. Failure is frequent and necessary in research, and is hard to get booted up.
If you don’t meet expectations over a long enough period, it might become evident that the research domain, your advisor, or some aspect of a PhD are not a good fit. Your advisor should have a discussion with you and clearly describe their thinking. In this case, you should still have the option to search out another advisor (inside or outside the department). This should really be the worst thing that your advisor can do to you: respectfully stating that they don’t believe the collaboration should continue. This must not be accompanied with threats, and they should be supportive of you moving on to your next endeavor.Development of skills. You should expect that your advisor provides you clear instruction of how you should develop your general skills as a researcher. This can include working with some infrastructure, practicing software development skills, presenting research, structuring your writing, and reading research papers. You should expect that your advisor will help with this by giving you opportunities for developing these skills, and provide input into how to manage your time.
Presuming that you respect your advisor, you should take their advice earnestly, and devote the time to developing these skills. However, you should always understand why you’re devoting your time to a skill, and be convinced it will be useful for you as a researcher and professional.
If there are additional skills that you believe you should have, you should feel welcome to ask your advisor for their input and, if they agree in the utility of the skills, how to acquire them. A PhD is about proving you’re a world leader in a domain which means devoting time to publishable research, but the skills you’ll acquire go well beyond that.Academic integrity. Academic communities are built up on trust. When you submit a research paper, it is peer reviewed, with the assumption that those reviews are unbiased and representative of a scientifically-grounded analysis of the research. If reviews are biased toward specific authors or institutions, the community suffers. Though reviewers do their best to analyze each paper, they usually do not attempt to reproduce the results, thus they rely on their intuition and experience to assess the validity of the results. There have been many stories about people gaming the system: friends providing biased reviews and accepting papers that might not stand on their own, papers that fabricate results to cover problems with the approach, etc.
Fundamentally, academia should not be a pursuit of glory. It must be a pursuit of knowledge. If enough of a research community departs from this focus, then there is no reason to put any utility in the results of the community. Thus, many communities have in place mechanisms to protect themselves. These take the form of lists of researchers whose papers must be rejected and researchers that are not allowed to be on program committees (thus provide peer review). These mechanisms can actually complicate the decision that a PhD researcher must make when told to fabricate results – they might feel like either choice (acting against their advisor, or fabricating results) will result in professional failure. Even in these cases, there are only difficult options, but if you want to be part of an academic community, you cannot hurt it by publishing fake/misleading results.Basic timeline, environment, and schedule. A researcher should have the ability to know what technical hurdles they must pass, what the timeline around those hurdles is, what equipment and space is provided for the researcher, Importantly, a researcher should know when/if their advisor has a sabbatical in the near future, and what is the plan for continued advising. Some institutions have a formal process about sharing this information, but my sense is that this is somewhat rare. An example (unfortunately through google translate) shows how useful something like this can be.
Summary. The PhD research/advisor relationship should be built on mutual respect, and should be based fundamentally on dialogue and choice. A PhD researcher should never feel like they have to do something that they don’t want to do. An advisor should be a mentor focusing on enabling the success of their students as researchers.
This is my brain-dump of the expectations for a PhD researcher. Am I missing any? Let me know on Twitter or in the comments below.
If you find yourself in a bad situation in which you’re being coerced, threatened, or generally not feeling supported, you have a number of options. I won’t argue that these options are easy, but it is very important to realize that you are not stuck.
It is possible that as you learn more about the problem domain and research, your interests diverge from your advisor’s. This is a natural, and somewhat common occurrence. Talking to your advisor, and your department chair should inform you of the options. However, if you’re in an unhealthy situation with your advisor, you might feel trapped, and it is understandable you wouldn’t want to discuss it with them. You should not feel like parting ways with your advisor is going to negatively impact your future, nor that you don’t have other choices.
One option is to talk to your department chair. You can ask for confidentiality in your comments, but you might want to ask them under what conditions can they not respect that confidentiality. Threats and harassment often have to be reported, thus breaking that confidentiality. You can get advice on how to handle the situation, but if you’re in a toxic situation with your advisor, the goal of this meeting is to understand the procedure for seeking out another advisor.
If you don’t want to be matched with another faculty, or in the worst case, your department chair takes the side of the professor and nicely tells you to “work it out with the professor”, then you need other options. First, you should always feel free to apply to other universities. Remember that there is a September and a January start time, though departments tend to accept more in September. The best way to ensure that you have options here is to ensure that you work on your communication skills, and on your network. To keep your options open, I suggest meeting and talking to other academics at conferences, and doing internships. These activities will give you a broader network of academics that you can leverage if you want to look for something new.
To protect yourself in case things go wrong with your advisor, you should focus on your communication skills. If you stay within your department, you might be asked to perform teaching duties, at least in the interim till you find another advisor. You should ensure that you work on your communication skills to the extent where you can be trusted to run a lab. To effectively talk about your research with a broad variety of researchers at conferences, you should work on your communication skills and ability to convey key points at a high-level. When doing internships, you have to work tightly with researchers with varying backgrounds, thus your ability to communicate your ideas and understand theirs is paramount. In short, to ensure that you have the maximum flexibility in your future, you should make improving your communication skills a primary goal. I have some notes on this.
Lastly, realize that if you were able to get into a PhD program, and have some amount of training in a deep technical area, you are competitive in the job market. Many students realize that a PhD is not for them, and often find rewarding jobs in which they are successful. Even if your transition away from your advisor makes it impossible to remain in academia, there are a monumentally large number of interesting problems that need solving in industry. Don’t build up your own self image solely around getting a PhD. Focus on doing interesting work, and realize that you’ll likely find it outside of academia if you must.
A number of venues2 for getting advice, help, and lodging complaints.
There are an innumerable number of problems with the academic system. Here I’m going to focus some of the pressures that cause the bad situations that many PhD researchers find themselves in, and on the power relationship that makes this problem is so hard to solve.
Some of the bad situations in academia happen because the underlying incentive structures encourage researchers to focus on quantity of publication. It is hard to evaluate a researcher’s work based on a qualitative evaluation (though it is necessary). If they aren’t in your field, how do you determine the quality, trajectory, and potential of their work? In reality, you can do a decent job at this evaluation, but it takes time and effort. What’s much easier than this? Counting publications. This factors into annual salary increases, department rankings, job offers, tenure decisions, acquiring grants, etc… Thus, there is an incentive to publish more, even if no-one reads the papers.
The side effects of this incentive system are varied. They include:
Most debilitating, I believe this focuses research on getting papers accepted, rather than on doing interesting research. Though accepted papers are often on interesting research, I firmly believe that motivations matter.
Put simply (quote):
Above all, quality and impact need to be incentivized over quantity. Sheer numbers of publications (or derivative bibliometrics) should not be a primary basis for hiring or promotion, because this does not encourage researchers to optimize for quality or impact.
Reasonable arguments have been made that in the world we live, you must focus on both quality and quantity. I certainly agree that the world does value quantity in addition to quality, and you cannot rely on those who evaluate you properly focusing (or being able to focus) on quality.
If professors focus primarily on continual publication as their primary goal, it is natural that pressure trickles down to PhD researchers. In this case, the entire focus is on deadlines and on getting results sufficient for a publication. In the LPU mindset, it is not useful to get results that are stronger than are necessary for acceptance, as you might as well use the additional results to get the next publication accepted. This dilutes the quality of each publication, and I believe this is wrong to the core. Not everyone agrees. Thus, it is important for a PhD researcher to ascertain what the core values of their lab are, and what the advisor values in research.
Students can easily feel trapped in a bad situation with their advisor. Many international students feel that the status of their VISA is under their advisor’s control. In the worst case, I’ve heard of advisor making physical threats. That the term “academic slavery” is something that many can recognize, if not identify with, is a sad indication of the situation.
A major issue is the advisor’s perceived lack of accountability, thus an inability of the PhD researcher to lodge a reasonable and serious complaint. If a PhD researcher believes that the department chair is unlikely to intercede, and that there aren’t outlets for complaint within the community, it is hard to see many options. Harmful advisors never need to change their behavior, thus will exploit PhD researchers across many academic cohorts.
At its core, academia is based on trust and responsibility. Thus professors are granted a large amount of independence, one of the main benefits of the job. Professor’s labs are similar to startups: small groups of individuals, aggressively pushing to get their technology adopted, while seeking a constant stream of funding. Professors are in charge of their research direction, and how they run their labs. This is by necessity as different research domains and areas of expertise require different directions and management. Further, they are supposed to co-manage the University and ensure it maintains appropriate research and educational standards. This leads to very shallow management hierarchies in which Professors are accountable to few. Professor’s management style and mentoring are under-evaluated during Tenure evaluations, but red-flags will be considered.
Regardless this independence, coercion, threats, sexual harassment, and the like, are incidents for which a Professor can suffer significant consequences. Unfortunately, the history of persecutions for such behavior leaves a lot to be desired. Regardless, PhD researchers in bad relationships are often less concerned with repercussions on their advisor, and more-so getting out of the negative situation. See the “Parting Ways” section above.
In short, professors have a lot of power over PhD researchers, and are often not that accountable. It is important to join a lab that has a healthy dynamic, and be willing to part ways if it isn’t a good match, or if the relationship is toxic.
In this post, I’m attempting to provide my perspective, given the current state of (American) academia. I also cannot more strongly recommend a number of current social movements including:
I don’t think that these are sufficient, but they are large steps in the right direction. Regardless, my focus here is not on how academia has to change (it does), but on what a PhD researcher should expect, and what do to if they are in a bad situation.
When you’re applying a position in a research group, you have almost no information about what academic life will hold, nor what the relationship will be like with your advisor. I want to provide a list of questions that might get you more information about the expectations and relationships of your prospective advisor.
It would be good to ask these questions to the current PhD students of the advisor, and compare the answers to that of the professor. In addition, you might ask the students:
If you’re an international student, do not feel like you have to work with a professor from the same country or region as you. Diversity is a core tenant of academia, and it is important to learn from various perspectives. Immersing yourself in a group from a different ethnicity and background will push you to vastly improve your communication skills.
There are quite a few times where I messed up as an advisor. These mistakes have made me reflect about the root problems, and clarify my parameters for interacting with PhD researchers. I share these mainly as examples of what I see as a legitimate set of problems due to a Professor making mistakes and struggling with constant improvement. I still make mistakes with an alarming frequency.
These shortcomings indicate that Professors are not perfect in managing their relationships. However, each of these have led to discussions about how the situation can be resolved for both me and the researcher. Importantly, I’ve reflected on how I can do better in the future.
I believe these are the scale of problems that are somewhat typical in academia. Professors aren’t trained in management, and we do a lot of learning on the job. However, problems with coercion, threats, and inappropriate pressure that removes PhD researcher choice, are not appropriate. I hope this document has provided some insight into the options that one has in these situations.
I use the term “PhD researcher” to mean a University student who is training for their Doctoral/PhD degree. I use this term instead of “PhD student” as it emphasizes the goal of the institution: to train capable researchers. Though individuals who are just starting the process might feel quite like students, the relationship with an advisor changes over time to peer researchers.↩
Please let me know of the ones I’m missing.↩