Advice to PhD Students

I got my PhD in Information Science at University of Washington, studying AI Education. Here’s 75 little bits of wisdom I picked up over a 6 year PhD process — hopefully things you won’t find in a generic guide for getting a PhD.

Applying 📝

1. Wherever you end up is the right place. I promise.
2. Hope for the absolute best for your peers going through interview loops with you. They deserve it too, and they’re on their own journey.
3. Listen to the PhD students when you’re coming in, but listen with discernment. No one can predict the future, or tell you how you’re gonna feel/fit in/suffer/succeed. They can only tell you their experience. Take what resonates and leave the rest. However, if they mention abuses of power or sexual harassment, take them seriously.
4. Have an idea and try to believe in it. But don’t be so rigid in your idea that you don’t seem teachable or flexible. Have your own project idea, but a willingness to be useful elsewhere. Read this beautiful children’s book, then read it again: What to Do With an Idea by Kobi Yamada.
5. The process of applying to anything is much more soul-searchy than I expected. Imagining myself in a new city, hoping I’m “smart” enough, good enough, wanted enough. Imagining whole trajectories of my life with every application; every rejection a reorganization of my hopes and plans. Treat your body like it is going through something, because it is.

That First Year 🤓

1. You really are smart, and you are doing everyone a disservice by trying to prove it all the time. You don’t have to be better than anyone to be here.
2. Some people have already published, started projects with their advisors, gone to your field’s major conferences. It probably helps them feel more confident to talk about it; let that be their story and don’t let it bother yours. You’re right on time. It’s the first year!
3. Your way of thinking is not the only way of thinking. You can use whatever worldview you want in your own papers; let your peers pursue their way of thinking. Maybe even listen once in a while. Their way of thinking is not a threat to yours.
4. Everyone will be giving you advice right now. Sometimes to hear themselves speak, sometimes to criticize you, sometimes to help you, sometimes to get you on their side, sometimes to extract your labor, sometimes as casual conversation, sometimes because they like you and don’t want you to get hurt. Take advice from people you respect. Take advice from the people who have something you would like to build in yourself. Learn from them, while honoring your experience and identities are your own.
5. You are special. And also you are not. You are absolutely brilliant, creative, insightful, talented, hardworking, and determined. You are also new, and you just haven’t had enough time yet to practice your skills. You’re in the right place! This is where you learn! It’s still school, afterall.

Classes 📚

1. There is no more school after a PhD, you’ve literally done all the school you can do. You do not need to get good grades. You need to apply your coursework to your research.
2. However, homework is one of the only “easy wins” you will get; when everything else is uncertain mush with maybe one acceptance a year, a good homework still deserves a gold star. Bring stickers. Give them to yourself. Maybe your statistics professor will laugh when he sees you putting a star on your own homework, mine did.
3. If you can, almost everything you do in your courses should relate to the project you’re working on. For me, my project was my dream. My classes were a tool to develop my dream. Not everyone approaches their first project like this. But I needed my vision to succeed: because in 2018 we didn’t know very much about responsible AI yet. And it felt horrible to see all the harm that these technologies were causing. I believed in my vision. I needed my vision.
4. Everyone is tired of hearing about your vision. Please let someone else speak.
5. You don’t need to be better than your peers, even if they are working on a similar problem. It is never exactly the same. Trying to be better than them doesn’t serve that dream of yours, it just makes you anxious and distracted and insecure. At the same time, you don’t have to hang out with the know-it-all who has a gazillion papers and a fancy tech job lined up after graduation. You can probably learn from them, but you really don’t have to. Just don’t be a jerk.

Insecurity 😬

1. If you can’t tell by now — I was very insecure.
   There were lots of reasons for this. I was switching from my Bachelor’s in Cognitive Science into Computer Science/Information Science, and felt behind. I wasn’t published yet, I didn’t know the conferences or the papers or the people everyone was talking about. I talked a lot, and most people found me quite annoying (I was). I had a looooong history of trauma and abuse that made me feel like if I wasn’t the smartest, I would be useless and thrown away. I was grappling with an autism diagnosis and my nonbinary gender identity, along with substance abuse issues all throughout my first couple years. Nothing could build my confidence other than time, but looking back I wish I’d trusted the process more: listened more, and talked less.
2. The PhD program is there for you to learn. If you were qualified already, you’d have a PhD given to you already by divine intervention or something. The program is also set up such that if you work on a project, listen to your advisor, read papers, collect data in a reasonable way, and write up your work as best you can — you will learn and succeed. I know it sounds too simple, but just show up. Just show up, keep going, you’ll get there.
3. Some people will be faster than you, more successful than you, more published than you, even liked more than you. That’s their journey, not yours. You are already getting a very prestigious and respected degree, even if it doesn’t always feel like that. Use this time to care deeply about your project: because it excites you, because it is useful, because it is teaching you a new skill, because it will help somebody. Focus on your own project, not someone else’s trajectory.
4. Sometimes you will be the one who is faster, more successful, more published, well-liked. Of course you’ve worked hard. You’ve also gotten a bit lucky. Teach others how you got what you have, so that they can be successful too. There is enough room for all of you. Practice this abundance mindset and you will see abundance everywhere.
5. Some days, my PhD work came naturally and I could learn things quickly and it felt easy. Other days, it felt crushingly difficult and I wondered how I would ever finish. While your path is entirely unique to your research and your results, there is also a formula to finish a PhD. Take your classes, take your qualifying exam, work on projects, propose your dissertation, write your dissertation. It’s not easy, but it is simple. You’re already here and doing the thing. If you want to finish, you will finish.

Building Software 🤖

1. I spent a lot of my PhD building small web-apps to collect data and test theories about how we teach AI and machine learning. I was not a software engineer, and I wrote horrific code. Before my PhD, I’d made cognitive science experiments (for adults, babies, and monkeys) using Python or JavaScript. No one ever taught me JavaScript. I published my first paper based on an interactive machine learning tutorial on gradient descent that I built using JavaScript. My JavaScript was very bad. Some of the data from that paper had hand-written responses because I didn’t know how to work a database yet. I had to transcribe the hand-written responses from the worksheets I gave my participants. The paper was still good, and it was the theories, hypotheses, and results that were interesting (not the app itself). Remember we are researchers, not companies.
2. The next software I made was a completely creative endeavor. What if we made an app where people could look at their own Facebook data? And learn about machine learning and algorithmic harms from that? That paper got in because of the “novel data collection method, despite the lack of strong findings”: aka, the opposite of the first paper. This time, my software was good but my results were kind of mediocre. Remember, if you have a compelling enough idea you can learn how to build it.
3. Well now I was fueled by these successes, and no longer felt the ego-serving need to build things entirely from scratch. APIs exist to do stuff that people already figured out. I was interested in how humans find biases in everyday AI technologies. So I made some R Shiny apps where learners could run their own images through Amazon Rekognition for emotion detection, or explore the NLP patterns of their favorite celebrities on Twitter. I did not know how to do any of that until I knew how to do all of that. It was still not very good. And that’s okay.
4. Don’t expose your API keys to github or in code you pass around to your peers. Your code can and will be garbage, but don’t push your credentials to git like I did when I was 19 (it turned out okay but I thought I was gonna die).
5. You will never know all the names of the doohickeys and the plumb bobs and the whatsits and the hoosits of software engineering. Every new thing I learn I feel like I have to upload to the SlapSmash after I ping someone on TimTam once I run my code through a FlimFlam on the YorkFork. You will figure it out as you go along, and you’re not stupid.

Collecting Data 📊

1. Begin with the end in mind: which columns will you compare? How will that test your hypothesis? What format will easily go into your R or Python script? If I want to compare one group to another, I need to have those groups linked to some kind of label of which group they belong to, along with the outcome measurement I’m comparing. Do I want that label in String format? Draw out the ideal dataframe that will get you your answer. Collect your data to look like that.
2. Collect a few things you think you might not need. A lot of data collection includes timestamps, region the data was collected from, etc. Keep all that, unless there is privacy reasons not to.
3. List out “sanity checks” and assertions for your data. Collecting age? Make sure no one is younger than 18, or 1000 years old. Make sure no one finished your task in under a minute, or took 2 weeks, or whatever it is. Running a script that computes probabilities? Better add up to 1. All the things that can possibly go wrong, get ahead of them. If one group is better than the other, do you expect them to be 50 times better, or a millisecond better? What is reasonable based on your expertise in the literature and area?
4. On that note, remember that we can make patterns in our minds about anything. We see patterns and we see what we want to see, because we want to publish our results. I liked to set up experiments such that even a null result told us something interesting. Map out the different results you could get, and what they might mean. As a PhD, you are not just a data processor. You have context, intuition, background, and reason. Don’t get too excited over a statistically significant result until you can explain what it really means.
5. And by the way, learn about effect sizes. In an A/B test I can test thousands and thousands of people using A or B. I can demonstrate, beyond the shadow of a statistical doubt, that A and B have statistically significant differences in behavior. It could be .0000000001 probability that such a difference occurred by chance. And that difference might be a fraction of a millisecond. Is that enough to tout option B as a cure-all? It might be, at scale. Who knows. But by understanding effect size you can make that judgment.

Writing Papers ✍️

1. Write the Introduction last. You’ll know what you’re talking about so much better after your analysis and background literature.
2. Write your methods first, while you are still doing them. Write everything down so you don’t forget, and also so you see those paragraphs appear in the document. It feels great and very official.
3. Write out your entire paper in 10 sentences or less. “Big problem” → “littler problem” → “way people have tried to solve littler problem already” → “way I think we could solve littler problem” → “how I test that idea” → “what we found” → “what it means” → “how it relates back to Big problem” → “what you should do next”. This helps frame the entire document, and now you get to just fill it in!
4. Submitting a paper is a success, no matter what the outcome is. The first rejections feel crushing. I even got a paper accepted and the feedback was still crushing. You will get feedback that makes the paper a lot better, and it will get in somewhere! I hated hearing this as a young PhD student. I’m still telling it to you.
5. You’re doing something really difficult, you should be really proud of yourself.

Publishing 🌟

1. People like to act like “publish or perish” isn’t real. They’re not a regular department, they’re a cool department, what’s the 411? Well here’s the thing. Publish or perish should never be a thing. But “publish or feel impending pressures and insecurity and disappointment in yourself and like you’re running out of time while everyone else around you succeeds and maybe your ideas are just bad and you should quit” is the reality.
2. I spent a lot of energy worrying about not publishing. I could have used that energy to work on my papers. Remember what I said earlier, the PhD is about showing up and putting in effort. Show up, write a paragraph. You’ll get there.
3. When you get a paper accepted, celebrate! You can go through the reviews another day. I used a system of highlighting “Positive”, “Negative”, and “Actionable” items.
4. When you get a paper rejected, celebrate! You can go through the reviews another day. I used a system of highlighting “Positive”, “Negative”, and “Actionable” items.
5. Publishing is one of the rare “big wins” in the PhD process. I personally wish I’d gotten sober earlier, and hadn’t spent so many nights celebrating my big wins with big alcohol. Instead, might I suggest a massage or a new book or a fancy dinner or several days of doing nothing in your pajamas!!! Whatever you do, you get a free pass for a while. You accomplished something huge.

Internships and Working in Industry 🤝

1. I know a lot of PhD students who had industry internships or did consulting for industry throughout their PhDs. Please know that if it doesn’t happen for you, it’s not because you aren’t smart. You will be well-served to work as a Research Assistant for a professor at your university as well, and will probably get a paper out of it! That paper may lead to industry positions. There are many paths to the same destination.
2. Yes, you are smart. No, you don’t know how industry works. Of course there are many types of industry jobs and internships. But I have heard from industry professionals that PhDs often come in thinking they know everything, only to perform poorly on the actual project. Come in humble and willing to learn. But be clear that you do have expertise you can offer, and carve out exactly how and where you can be useful.
3. You may have to think about revenue and costs when it comes to your ideas. Code that works for your data collection might cost thousands when scaled to more people. Whereas you might get eager unpaid undergrads (questionable) to label your data or test your software, the company has to actually pay people to do those tasks. Choose the tasks wisely, with the best chance of returning value.
4. A theory that is interesting may not translate to an actual thing that works. A paper you read that feels relevant is only relevant if it solves the problem. Possibilities that excite you should be acted on, not recommended at the end of a paper after your limitations. And when it comes to limitations, some of those won’t fly at an industry scale.
5. I’m making all of this sound like there aren’t wonderful and curious people in industry; there are. Not everyone is the Monopoly Guy trying to take baths in bags of money; most are not. If your principles or values are challenged, remember that you are the expert in the room. Tell them what they need to do and why. Don’t compromise your ethics, but do communicate them.

Conferences 💬

1. I remember my first conference I would “ask a question” every chance I got. I was not asking a question, I was advertising my own work 😅. Don’t do that. It is annoying.
2. However, you really can ask questions to help you succeed in your own work! “How did you do X”, “How did you handle Y”, “Where do you think the next steps are in this area?”. You throw the presenter an easy question while also validating your own intuition or learning a new avenue. Everybody wins.
3. If you’re presenting, have fun! You already published the paper! Share what excites you about it, and don’t think of it like a test. You already passed the test, this is the celebration.
4. If your conference does not have a Sensory Room or other accessibility accommodations, tell someone. Many conferences are taking care of this on their own right now, but many others are simply waiting for someone to tell them how to do it. For example, a Sensory Room can quite literally be a piece of paper on a lowly lit room that says “Sensory Room” or “Desensitization Space”: it usually does not cost anyone extra.
5. You cannot and should not attend everything. Even if it’s a networking opportunity. Even if it’s really cool. Even if it’s related to your work. Yep, that’s right. Pick ONE non-negotiable for each day of the conference, and schedule your breaks. Explore the city. Be done when you want to be. Networking can be great, but everyone is so overstimulated at this point that you should prioritize the opportunities that really matter, and make time to enjoy yourself the rest.

Identity 🌱

1. Hate it or love it, you are building a brand for yourself throughout your PhD. In my experience, the best way to do that is to be unabashedly yourself, and to genuinely like who you are. You can’t “trick” anyone into thinking you’re anything that you aren’t. The good news is, you don’t have to. You are the asset. Your exact distribution of skills is like no one else’s.
2. It took me years of trauma recovery therapy to learn to even like myself. I had to untangle childhood abuse, a pervasive eating disorder, my reliance on substances, a confusing nonbinary identity, and a late autism diagnosis. In the end, what I needed to heal was a desperate need for validation from others.
   I hope it is easier for you than that. What got you this far is so much talent, determination, creativity, grit, and sparkle. You have done enough. You are doing enough.
3. It could absolutely be the case that your identity is bigger than your research. For me, I shaped my research around my identities. If my research didn’t match up with something I really believed in, I felt unmoored. The beauty of a PhD program is you have the freedom to explore your ideas and what you care about. Try your best not to waste that freedom. My work has explored topics like disability, self-advocacy, eating disorders, sexual assault, trauma recovery, gender, anti-recidivism, substance abuse, mental health, and political protest: all within the bounds of studying AI and computing education. I am so lucky I got to do that.
4. You may be asked to be on diversity initiatives, often unpaid. Think carefully about this, and the precedent that it sets. You may jump at the chance because you want to be known by a professor or the Dean or the department. You may think it will bring you accolades. It might! It also might bring you a bigger headache than you anticipated. Ask questions, ask why it is unpaid, ask what the clear outcomes are, ask who is there to support you, and ask how your committee or initiative will be put into action etc. You are valuable, and so are your peers. When people pay for things, they also invest in their success. Consider a small honorarium (whatever they pay guest speakers).
5. You don’t have to be constantly branding/selling yourself. Make a website, write out your mission, and try to talk to people. Remember to do the footwork of the PhD program, and trust the process.

Teaching 🎒

1. In my experience, your students either want to get an A, or get a job, or both. Make it as clear as possible how to get an A, with a clear upfront schedule and expectations and rhythm. Rhythm of a course means similar formats between assignments, similar amounts of effort needed for those assignments, and clear expectations that are not violated halfway through. Everyone benefits from this.
2. If your students are aiming to get a job, tell them you understand this. Tell them you understand how much pressure they must be feeling — and that your coursework has been created with a career in mind. When they know you are on their side and understand their financial pressures, they work with you not against you.
3. Course evaluations reveal some nasty attitudes and comments, especially from students who did poorly or disliked you. Course evals are often a place for students to feel a sense of agency and power, something they do not always get to feel. Let them vent, try not to take it personally, and highlight the good comments. Even better, have a friend filter them for you first.
4. As an educator, you are training the future leaders of the world. Help them thrive and believe in themselves. They will remember you.
5. Students cannot engage in your course content if their material or emotional needs aren’t being met. Be curious about your student’s material and emotional safety first. Then teach the math.

The Unexpected ❤️‍🩹

1. In my second year, I testified with a group of four others in an unexpected legal case regarding my childhood SA. The pandemic hit a few months later. I had just quit my lab with no fallback. I developed a drinking problem. I spiraled down down down for 2 years until I took control of my life back. I got better. I fell in love. He died of a fentanyl overdose six months before I defended my dissertation. Remember that life happens during your PhD, and you are human first. I talk about these things so you know you are not alone. Your peers are not robots; they are going through their own tragedies and hardships. Lean on each other.
2. You will not function at the same level as you did before tragedy struck. That is okay. You are still brilliant. You are still worthy. Let yourself heal.
3. While you may not be able to control the future, you can set yourself up for success. Go to the doctor. Get a therapist. Drink more water than coffee. Sleep at a reasonable time. You’re kind of in college but you’re really not in college. Take care of yourself.
4. The unexpected will guide you to the next step. Without the love and death of my partner, I wouldn’t have gotten CPR and Narcan certified, nor would I have had the opportunity to help out a nonprofit on their overdose awareness training and data collection. I wouldn’t have developed such a deeply personal interest in anti-recidivism or overdose awareness. And I wouldn’t have seen all the rich humanity and empathy in all my wonderful peers and colleagues who reached out to help. I wouldn’t have developed a deeply personal spirituality that keeps me connected to nature and the flow of things. None of it is better than having him back would be, but I accept life on life’s terms now.
5. You will survive this.

Mental Health 💓

1. 5+ years is a long time to be working towards a goal, and a lot happens in those 5+ years. By the time you get your PhD, you will not be the same person you were when you started. Heartbreak, tragedy, stress, disappointment, life will happen. You are human first.
2. Very rarely is there an “emergency” in your PhD work. I only worked in Biology labs for a little while, and I know things can definitely feel urgent there. Data breaches and exposed information is an emergency. But almost everything else is not as urgent as it seems. Your PhD work can hold off for you to feel better.
3. There are a gazillion resources for mental health on university campuses. But you are the only one who can determine what is actually right for you. Therapy is still stigmatized in some communities. Psychiatric medications even more so. There is usually no quick fix for mental health, and you are the expert in your own experience.
4. One of the things I need for my mental health is to feel purposeful. And when a paper has been written and rewritten four times and it keeps getting rejected and you don’t care anymore, it’s hard to feel purposeful. This is a great time to turn to the overarching arc of your dissertation work and why it matters to you and the world. I called it my pursuit of my “sacred work”, because I am dramatic.
5. Just another reminder that you are, in fact, good enough. Keep going.

Your Future💡

1. I’m writing this after graduation but before a full-time position. In the last couple months I taught myself a software engineering stack, caught up on AI policy, dreamed up a children’s book, made art, started a consulting company, and worked daily on my self-confidence. Your future is bright.
2. Your dissertation matters, and may set the foundation for many things. But it is just one small piece of your career. You cannot and will not say everything you wanted to say in your dissertation. That leaves room to say and do those things in your future! A good dissertation is a done dissertation.
3. Take time to celebrate how far you’ve come. Remember how it felt when you were applying to PhD programs; how much you hoped and prayed and worried. You did it. You got to have an experience that very few people have, and you will use everything you gained to change the world and inspire others. Hang up your diploma, it feels very official.
4. There is no ceiling to what you can do or who you can become. Someone recently told me that, even though I struggle with self-confidence still. I’ve been repeating it to myself as often as possible. Let’s say it again: There is no ceiling to what you can do or who you can become.
5. Congratulations, Dr! Now you can pass on your insight and inspiration to younger yous. You certainly don’t need my advice anymore.