Essays

Not a typical essay, I recently shared this document with the Stanford CS department as a guide for recruiting and supporting undergraduate researchers:

Inclusive Recruitment and Support of Undergrad Researchers

Overview

  • Providing ample opportunities for students to get involved in research as undergraduate students (and supporting them once they are!) is one of the most important ways to foster diverse talent and remove barriers to research-based careers

  • But it can be challenging to do this well, especially outside of formally-structured research internship programs.

  • This guide is intended to offer some guidance for grad students and faculty towards this goal. These tips were compiled based on suggestions from undergrad mentees and grad student/faculty mentors at Stanford.

  • Please feel free to share, and reach out to Alex Tamkin (atamkin@stanford.edu) with any questions, thoughts, or feedback.

Advertising: best practices

  • Cast a wide net

    • Don’t recruit solely via friends of friends, word of mouth, or waiting for students to reach out to you

      • This severely limits the distribution of students you will consider

    • Do write up a job posting + distribute it broadly

      • E.g. there are email lists that go to all students of a certain major (consider a wide range of majors as is applicable to the position)

      • Make sure your posting is listed on any boards/forums for listing research opportunities

    • Do send (judiciously) to a diverse range of undergraduate listservs. Be respectful of each mailing list's rules about sharing opportunities. Some ideas include mailing lists for…

      • Different student communities

      • STEM student groups

      • Present or former section leaders

      • Students considering a certain major (e.g. considering-cs)

  • Advertise early, not last minute

    • A couple of weeks before the quarter/semester (or earlier if possible)

      • This makes it easier for folks to plan, especially students who have to work jobs + figure out things in advance.

      • Also tends to result in a more diverse mix of students

    • Info session / panel / "office hours" can also be effective (either for individual projects or if multiple positions are opening for a lab)

  • List a timeline

    • Deadline for students to apply, and when they will hear back

    • Example:

      • Application deadline: 10/1 (apply via this form: <link>)

      • Interview notifications: 10/8

      • Decisions: 10/15

  • Set expectations

    • How much time do they need to commit per week?

    • How often will you meet with them?

    • Will they need to take research credits?

    • Whether or not funding is available

    • What level of authorship can they expect?

  • Tip: Intake with a survey form rather than email, to gather all the information in one place

    • Consider asking for resume, classes they've taken, and other information if you believe it's relevant

      • But: focusing too much on prior experience, especially when not strictly necessary, can overly winnow down the pool (mentors have found that other factors like interest / motivation are often more important than past experience)

      • Consider stating that students should still apply even if they feel they don't have enough experience (you can make that assessment yourself)

    • Consider not asking for gender, race, & other demographic information. If you must, consider asking for it at the end. Some studies suggest that underrepresented minorities lose confidence after indicating their demographics (stereotype threat), which may impact their survey answers. However further study might be necessary to improve confidence in the replicability of these findings.

Writing up a job posting

  • Scope the way students could be involved based on your available time and goals. There are multiple setups, including:

    • A) Individual mentorship

      • At some schools, it's common to treat undergrads as if they were mini PhD students, giving them significant independent responsibilities

      • This can work well, but often requires more work from mentors, meaning they can only take on one or two students

      • This contributes to a significant scarcity of undergraduate research positions, meaning that only a small fraction of students (often the most prepared) can get spots

      • This can still be a good model, but consider casting a wide net

    • B) Group mentorship

      • There's a much larger group of motivated, prepared undergrads who can play strong contributing roles to a research project

      • Often 2-3 students can join a project + associated meetings

      • Can lead to faster progress, give students an opportunity to learn from each other, and help remedy severe shortage of research positions—a potential win-win-win

      • But while group settings have benefits, they also have certain downsides: students may have different amounts of experience, have different working preferences, or feel an implicit sense of competition or comparison. It also may require a different set of management skills vs mentoring one student. You'll have to decide what makes sense based on your experience and research area.

    • C) Research Immersion

      • Undergrads don't have to do significant contributing work to a project—they can also just observe lab meetings, help with simpler scripts, data processing

      • This gives students cultural exposure to a research environment (can be useful information for student decision-making about longer involvements), without requiring too much mentor overhead

      • Don't need to gate involvement on authorship

      • Note that this has to be a two-way street—clearly advertised so students can choose if this is the kind of experience that they want

  • Project descriptions

    • Especially if you're hoping to reach more junior students, it's helpful to putting a bit more detail about potential projects in your posting

    • Try using some simpler, non-technical language, as this may be more likely to elicit interest than more terse, technical buzzwords

      • E.g. elaborate "multimodality" with "understanding the relationships between different kinds of data, such as images and text"

      • Make clear what's necessary to know beforehand vs what can be learned on the job—don't want folks self-selecting out even before they apply

    • Consider providing links to related resources for students to learn more about the topics

  • Requirements

    • List specific courses

      • Avoid generic "Coursework in AI / ML"

      • List specific courses, e.g. "one of 221, 224n, 231N"

      • List the minimum number of courses necessary

    • Avoid generic subjective descriptors

      • E.g. "Ability to pick up new technologies in CS quickly"

      • Not likely to get you who you want, may unnecessarily filter out people who would be good fits but are less confident (disproportionate impact across demographics)

    • Consider adding a note like the following, which was perceived as a welcoming / good sign by students

      • "If you are unsure about whether you have the appropriate background, feel free to send us an email and we can help you."

  • Consider not limiting to certain years

    • Many students are quite capable and ready to get involved freshman / sophomore year (and bring lots of energy!)

  • Consider setting up brief office hours (e.g. a couple short sessions) during the application process

    • Let students ask you about their applications and for advice

    • Builds connections, and supports students who might not feel comfortable applying otherwise for whatever reason

Selecting students

  • Don't count students out because of outside interests / activities

    • As surprising as it may seem, we've heard of students being passed over in other institutions because they have non-STEM outside interests, and this was (wrongly) considered a negative indicator of their quality as a candidate

  • Do consider a student's bandwidth and be clear of the commitments (time + otherwise) you expect from them.

    • Consider asking them to formally commit to these before beginning work in the lab—this helps make very clear the level of involvement entailed by the position, and minimizes the chance of miscommunications or students dropping off midway through the quarter

    • Also explain what "counts" as research—some of the time will be spent reading papers, self-studying, and brainstorming. Not all meetings may have a deliverable coding / theory update.

  • When considering students with little CS experience (e.g. first and second years) useful things to consider include:

    • How their other experiences and skills may enable them to contribute to the project/lab

    • Their expressed and demonstrated motivation / interest

    • Their potential for growth in the lab, including their availability to work for a long enough period to enable that growth

    • Supplementary sources of mentoring or support (e.g. CS 197 at Stanford)

Supporting and retaining students

  • It's not enough to recruit students—you have to support them after they join so they have a good experience and learn/grow!

  • Make it clear that not knowing everything (or most things) is really common, and that asking questions is ok! Students can easily become intimidated.

  • Explain how research is different than schoolwork in that things won't have correct or known answers, and being confused a lot is ok

  • Spending some time doing pair coding or other kinds of research together with your students can be a high-leverage use of time—building useful skills + rapport, making research progress, and helping students feel less lost

  • Leave space in meetings to talk about non-research stuff—express your care about your students' wellbeing

  • It's important to make students feel like they're part of a community / larger group. Include students in lab meetings, social events, and other group functions as possible, and facilitate fostering personal connection with graduate students, faculty, etc.

  • There's much more to retention than just these points—see this report, especially the summary section "Promising Interventions"

Acknowledgments

Thanks to Ahmed Ahmed for compiling a list of pain points from student discussions as CURIS outreach mentor, which served as an impetus for this document. Thanks also to Ahmed, Lisa Yan, Breauna Spencer, Megha Srivastava, Shreya Shankar, Isabel Gallegos, Stephanie Tena-Meza, and Keith Schwarz for helpful guidance, feedback, and conversations!