Faculty/Group Assimilation

Yue Crew

(Insert Date)

Objectives

A check-in on questions, expectations & concerns which need to be clarified

The group would like to understand

• Style of running group
• Style of doing research
• Style of communication
• Any changes to these styles

Objectives

Promotes group harmony by surfacing issues and proposing constructive changes

Develop comfort and sets a climate of openness

Ensures organizational effectiveness

Additional Incentives:

• Group members learn more about themselves
• Strengthens relationships

Timeline

Session 1: Data Generation & Data Review (2 hours)

• Does not include Professor
• Summary of feedback/questions sent to Professor

​

Session 2: Data Discussion & Next Steps (2 hours)

• Professor responds to feedback/questions from Session 1

​

Ground Rules for Session 1

Everyone has opportunity to speak, but don’t speak over each other

Speak openly, can re-frame wording later

What is said in the session stays in the session

Truly a brainstorm

Feedback is meant as a “gift”

Outline of questions

1. What do you know about your Professor?
2. What do you not know but want to know? (don’t get too personal)
3. What is working well and should not be changed
4. What isn’t working well and should be changed
5. What expectations do you have of your Professor?
6. What expectations should your Professor have of the group?
7. What is working well or not working well in terms of group dynamics?
8. Are there any immediate changes/decisions for your Professor?
9. How would you like ongoing feedback?
10. What is your Professor doing right?

​

• Action items (only for Session 2, done collectively by group & professor)

Example Student Feedback/Questions

& Faculty Responses

Example group feedback/questions in red

​

Example faculty responses in blue

​

Q1: What do you know about your Professor?

• He likes video games
• He likes basketball
• He sings in a choir
• His name means lone tree on mountaintop / pine tree on mountain
• He has a little sister, adult, lives on the East coast
• He has 2 super cute kittens

Yisong: I typically like to add a few photos

Q2: What do you not know but would like to know?

• What works are you most/least proud of?
• What is the group’s mission?
• What are your 3, 5, and 10-year visions for your research group?
• What is your favorite fiction book?
• Where do you wanna travel after pandemic?
• Have you ever had imposter syndrome?
• What does Yisong dislike about the ML community culture?
• What characteristic is Yisong is most proud of / dislikes in his students (in general, and not individually)?
• What aspects of your job have changed between the start of your assistant professorship and now?
• Is there anything that you’re trying to implicitly encourage in the group culture/social dynamic?

(A few example responses in subsequent slides, professor should address all questions)

What works is Yisong most / least proud of

​

The most fulfilling personal journey is seeing my thesis work have impact in a broad range of “recommender” systems that include clinical therapies.

The most exciting is typically the next cool result someone from the group discovers that is surprising and thought provoking.

The least proud are flawed projects that I rushed and am now embarrassed to have my name attached to.

What is the group’s mission?

Apply machine learning to the hardest problems we can (and in a way that matters), study why it breaks, and ultimately understand machine learning better.

Caveats & Notes:

• Different students focus more on different aspects (e.g., some students are more theoretical, some more focused on applications of AI for Autonomy/Science, etc.)
• Should try to cover this entire spectrum to some degree before graduating

​

What characteristic is Yisong is most proud of / dislikes in his students (in general, and not individually)?

​

Proud of: Being driven by genuine curiosity and asking big questions.

Dislike: Not much, really. I think I was less patient when I was a new professor, but now I’ve learned that each student needs to find themselves & their research vision at their own pace (and it’s also a multi-faceted process), and I’m mostly content to help nurture that along rather than force things. (I realize this answer is somewhat controversial given the later discussion items.)

Q3: What is working well and should not be changed?

• He is adaptive - adapts to what students want & need (i.e. number of meeting times, etc.)
• He connects you well with external collaborators, and collaborates well with other professors
• He gives good feedback on the PhD progress & milestones & student presentations
• He gives important feedback when it is needed and is kind about it
• The send-offs for students who are leaving Caltech are good
• He facilitates student meetings, such as the ML for science sub-group meetings
• He is supportive for student going on internships

(Typically no faculty response needed to this question.)

Q4: What isn’t working well and should be changed?

• He could do a better job with high-level vision, not just individual papers, but the collection of papers for your PhD
• Need for both formal and informal types of interactions. Need less formal settings, or meetings/short seminar intended to introduce you to the basics of a topic
• Facilitate awareness of what other people are working on and have expertise in. An atmosphere where anyone asks questions to the rest of the group about whatever they’re working on/need help with.
• More funding for social events / more clear funding
• (There is disagreement on this point) One thought, maybe he could be too adaptive (for example, early on during the PhD, a student might need more structure until they reach an equilibrium on the frequency of meetings)
• He could be more critical of our work - we want him to make us better.
• Giving more explicit feedback on presentation style, e.g. how to bring people up to speed in the 1st part of the talk. Foster people giving each other feedback.

(A few example responses in subsequent slides, professor should address all comments)

He could do a better job with high-level vision, not just individual papers, but the collection of papers for your PhD

I don’t think it’s my place to impose a high-level vision on to any student.

I view my role as helping the student sharpen their own high-level vision.

The main place for a student to start practicing is a thesis proposal (so the sooner the better for that).

Generally speaking, the foundation of a thesis proposal is your first “main” result, and the thesis grows from reflecting on the implications of that result and extrapolating.

He could be more critical of our work - we want him to make us better.

​

“More critical” can mean a few different things.

Some scattered thoughts:

• I prefer to criticize at the level that is “useful”. I think of providing feedback like gradient descent: what is the most useful local feedback given where the student is at?
• I think it’s hard to think about research vision in a meaningful way without mastering the process of doing a single research project
• I think it’s better to be constructive and prescriptive (which can be actually quite critical without sounding critical)

Additional feedback welcome.

Giving more explicit feedback on presentation style, e.g. how to bring people up to speed in the 1st part of the talk. Foster people giving each other feedback.

• This was an unfortunate casualty of my responsibilities skyrocketing over the past couple of years. I think creating a group-support system for feedback would be great. I had this as a PhD student at Cornell w/ other PhD students there.

​

• Brainstorming ideas:
• how to give general audience talks, how to motivate the research questions, give examples,
• give each other feedback, non-technical feedback (allow people to leave after an hour)
• explicitly stating the purpose of giving a talk

Q5: What expectations do you have of your professor?

• Guidance through PhD / developing research vision
• Career development: connect us to other people, help us reach our goals
• Help you evaluate all the things that contribute to your research process that isn’t easily measured by output such as papers, and evaluate the things that you have to do behind the scenes
• Do a better job of promoting the lab on social media, and the individual members on social media
• Yisong: group social media account made!
• Create a strong sense of “you’re in the lab, you belong here.” When you’re a new student and don’t have an advisor, you feel a lot of pressure to prove yourself, and having a sense that you belong can help you settle in. Having a sense of inclusion and that everyone belongs equally, for newer and not-so-new people as well. Treat students fairly, etc.
• Have our back, e.g. emotional things, medical problems, etc. Value the totality of our experiences as grad students + postdocs
• Organize research group meetings at some recurring schedule (1 hour, 1x a week)
• Research meetings (internal, about people’s projects at different stages)
• Tutorials (for different sub-groups, ex: ML & control, ML for science, …… )
• Yisong: high-priority action item

(Most comments don’t require faculty response)

Q6: What expectations should your professor have of the group?

• He should expect us to be proactive (arranging meetings, forming collaborations, doing research, finding papers, ...)
• We should be a stronger unit: promote each other’s work, if we see a paper that’s relevant to someone else, then forward it to them. Being more connected and helping each other out. Be supportive of each other.
• He should expect us to be honest with him about what we need from him
• Agency: we should have a vision for ourselves on which we’re trying to iterate. Have our own plans, not expect Yisong to exclusively have a vision for us.
• Yisong: generally speaking, I think this is great for 3rd year (and older) students. However, for early-year students, it’s usually not so important to be so specific.
• Group members should be able to influence the professor's research direction, based on what they think is promising, impactful. Emphasize future work/directions in lab meetings; could even have a whole lab meeting on this topic. Discuss what should be done as follow-up work

(Most comments don’t require faculty response)

Q7: What is working well or not working well in terms of group dynamics?

• Everyone from the group is very nice & approachable.
• We need a functional social chair, that cares about organizing activities
• Yisong: to be discussed in action items
• People are respectful about each other’s research
• Each person’s research direction is complementary - we learn a lot from other people’s group
• We should leverage these diverse backgrounds more
• Need a better understanding of everyone’s expertise (ex: control, optimization, etc.) - would be good if we had some sort of centralized source of this.
• Yisong: To be discussed in action items
• Often times we come up with new ideas, but there may not be follow-up on the ideas. We need to hold each other accountable on ideas
• Yisong: I think it’s OK for ~80% of ideas to have no follow-up.

(Professor should directly address things that are not working well)

Q8: Are there any immediate changes/decisions for your Professor?

• Cultivate cohesion. We should be able to connect with each other independently of Yisong or anyone else.
• We should be empowered to make the changes we want to see, e.g. random lunch assignments.
• We should have a conversation about social media strategy for the group.
• How much funding is available for social activities?
• Yisong: it’s a bit dependent on available funding, but I’ll try to come up with a more standardized framework.
• Sometimes when talking to Yisong about research, he seems distracted. He checks his phone in 1-on-1s and group hangouts; we’d appreciate it if he didn’t.
• Yisong: sorry about that, I’ll minimize distractions.

(Most professor responses can be deferred until action items at the end)

Q9: How would you like ongoing feedback?

• Interpretation: closed-loop feedback - between students & prof
• For group: We like continued group assimilation
• This format is nice because the whole group is together without Yisong’s presence.
• This is a good frequency (once per year)
• The feeling is that Yisong did a good job responding to the feedback from last year
• We got a group Slack that people used (yay!!!)
• Group twitter is also going :))
• New people should be particularly encouraged to give feedback on what they think, especially as they have a new perspective and have a higher stake in the group dynamic

(Most comments here do not require specific professor response)

Q10: What is your professor doing right?

• Yisong is a approachable
• Feels like we can go to him with our problems if we need to
• Yisong is accepting of good work-life balances
• Maybe express it more to students
• Yisong cares about representation & diversity
• Yisong is responsive and shows up on-time to meetings; if he’s going to be late, he lets you know
• Yisong really allows you to push your own projects, instead of trying to control what you’re doing
• Yisong encourages people in the group to work together
• Yisong is open to change - very open to new fields of research
• Yisong gives good suggestions on writing papers

Action Items

Provide better guidance about crafting a research vision

• Maybe more wisdom sharing by senior students?

Be more proactive about providing critical feedback

• Balance against other needs (don’t overwhelm student, signalling uncertainty so students don’t overfit to feedback, etc.)

Establish more standardized framework for funding (social & compute/hardware)

Set up a more decentralized & regular group meeting structure. Weekly? With Food?

• Social events, selected topics reading groups, etc.
• Establish group peer support structure.

Restructure group meetings so that more students get something useful out of it

• Advice sessions, thinking through steps in PhD
• Advice on good writing, presentation, graphics/visuals, social media, etc.
• Introductory tutorials on technical topics