How to run a high-energy reading group

Why are reading groups and journal clubs bad so often?

I think there are two reasons: boring readings and low-energy discussions. This post is about how to avoid those pitfalls.

The problem

I have participated in (and organized) some really bad reading groups. This is a shame, because I love a good reading group. They cause me to read more things and read them more carefully. A great group discussion will give me way more than I’d get just by taking notes on a reading.

This is what a bad reading group looks like: six people gather around a table. Two kind of skimmed the reading, and two didn’t read it at all. No one knows quite what to talk about. Someone ventures a, “so, what surprised you about the paper?” Another person flips through their notes, scanning for a possible answer. Most people stay quiet. No one leaves the table feeling excited about the reading or about being a part of the group.

This is avoidable, but you need to find interesting and valuable readings and you need to structure your group to encourage high-energy discussions.

How to find good readings

If you’re lucky, someone in your group will propose a reading that they’re excited to re-read and discuss in depth. However, at our East Bay Biosecurity meetups, we often wanted to learn about a topic (say, "regulation of gene drives" or "basic immunology") that no one in the group knew much about. A Google search for “basic immunology biosecurity” will not reliably find interesting readings. What are better ways to find good readings?

1. Follow a syllabus or reading list

People may have already compiled good readings on the topic you’re interested in. Take advantage of their work! Some examples:

EA syllabi and teaching materials
Annotated Bibliography of Recommended Materials from the Center for Human-Compatible AI at UC Berkeley
My Biosecurity and Biorisk Reading+ List
Most of the 80,000 Hours podcasts include links to recommended further reading

2. Ask an expert for recommendations

Find someone who is working on the problem you’re interested in and ask them for advice. It’s okay to cold-email people, and to send them a follow-up or two if they don’t respond (see It Is Your Responsibility to Follow Up).

You’ll get better responses if you give some (brief) details on your interests and level of background knowledge; I’ve received random LinkedIn messages asking “how to learn about biosecurity” and that’s not enough information for me to give useful recommendations. A (hypothetical) better example would be something like, “I’m running a reading group of undergrads (a mix of life sciences and computer science) and we’re currently trying to understand how to improve vaccine availability in future pandemics. I reached out because I saw you were involved in [thing]. Are there any papers or readings you’d recommend for the group? Thanks so much for your time!”

3. Browse back catalogs

If you can identify a few organizations, researchers, or journalists whose writing you enjoy, just read their work for a while. For example, in East Bay Biosecurity we read a lot of reports from the Johns Hopkins Center for Health Security and the US National Academies Press.

4. Ask one group member to identify good readings by giving a talk

If you want to learn about a topic, but you can’t find anything useful from a syllabus, expert, or back catalog, someone is going to have to wade through messy search results until they find something good. To encourage productive Google-wading, I suggest you nominate one of your group members to give a talk on the topic of interest. Preparing the talk will force them to read many things about the topic, and they’re likely to find useful standalone readings along the way.

This will require quite a lot of time on that one group member’s part (I recall East Bay Biosecurity members spending between 4 and 20 hours preparing short talks) but they’ll walk away with a strong understanding of the topic, and there’s no risk of asking everyone in the group to spend time on a bad reading.

Four structures for high-quality discussions

A good reading doesn’t guarantee you’ll have a good discussion. Will enough people read it before your meeting? What if everyone reads it, but no one has interesting questions to discuss?

Below are four models that I’ve found lead to high-quality discussions.The models differ on how much pre-reading commitment is expected from each group member and how heavily the discussion is structured. The way I organized reading groups for East Bay Biosecurity (vaguely hoping people would do the reading and spending ten minutes right before the meeting frantically scribbling down questions) is, uh, notably absent.

Table with four quadrants that list the four types of reading group below; read together, one person presents, two-person group, everyone has a role

High Reading Commitment, Low Discussion Structure: Two-Person Group

In a two-person reading group, you cannot hide behind anyone else. If one of you doesn’t do the reading, you have to cancel your meeting. This social accountability is highly motivating!

You don’t need to worry too much about structuring a two-person discussion. Just have a conversation. I have found “flip through the paper or book and pick out anything I had highlighted” to work just fine.

High Reading Commitment, High Discussion Structure: Everyone Has A Role

Maybe you want to include more than two people in your group. If you want most of them to do the reading, I suggest giving them a role! Here is a list of roles I am using in a reading group at the moment:

Discussion Generator: come up with 1 or 2 questions for the group to discuss
Summarizer: prepare a 3 to 5 minute summary of the reading
Highlighter: pick 1 or 2 passages that you think are great and merit further discussion
Concept Enricher: pick 1 or 2 words or concepts you feel confused about and do a bit of research on them, reporting back on what you learn (e.g. “gain-of-function”, “TET Enzymes”)
Connector: share 1 way you might apply ideas from the reading in your own life or work

You have to do at least some of the reading to fulfill these roles, which creates accountability. I derived them from my grade-school experience with Literature Circles, but omitted some of the more playful roles (e.g. an Illustrator who makes a piece of art related to the reading). You can add those back in if you like! For another Literature Circles adaptation, this post describes roles for a software study group.

The (highly structured) agenda for your reading group meeting naturally becomes cycling between people, something like:

Summarizer
Discussion Generator
Highlighter
Discussion Generator
Vocabulary Enricher
Discussion Generator
Connector

In my experience, this fairly lightweight structure avoids awkward silences and boring tangents, since you want to make sure to get to everyone’s contribution during your meeting time. This structure won’t work if you don’t know ahead of time who is going to attend your meetings or if you want to include more than half a dozen people in your group. In those cases, I recommend the next structure on my list.

Low Reading Commitment, High Discussion Structure: One Person Presents

This is the structure I’ve most often seen in academic journal clubs. One person is responsible for presenting the reading to the group and facilitating discussion. It’s common because it works!

In my experience, most members of this kind of group will skim through the reading’s section headings, but few will read every paragraph. So the presenter needs to start by summarizing the reading. In addition, they should look for interesting critiques of it, contextualize it in the broader literature, and spur discussion amongst the group. Basically, the presenter is giving a short seminar about the reading. It’s a lot of work for them, but each group member should only be responsible for presenting every few months.

This structure lets you include a larger number of people in the reading group (I’d say a maximum of 15 or so) since group members who aren’t presenting have a low reading time commitment and the discussion is structured and managed by the presenter. The presenter can also critique and contextualize the readings in more depth than is feasible in other structures.

Low Reading Commitment, Low Discussion Structure: Read Together

The most radical version of low reading commitment: do the reading while you’re all together as a group.

This structure is moderately fashionable in tech management, where it’s called Silent Meetings. Years ago, Amazon declared that Powerpoint presentations would no longer be allowed in meetings, and instead people would prepare 6-page memos to present their ideas. Since no one does pre-reads, meetings would begin with everyone silently reading the memo and making their own comments on it.

For long, dense documents (i.e. most of the things your reading group is likely interested in) I wouldn’t recommend sitting silently and reading the whole thing through. It would take too long. Instead, I suggest reading sections or specific pages of the document in silence for 5 or 10 minutes, then pausing to discuss anything you found interesting or confusing. If someone finishes the section before the time is up, they should use the time to either silently take notes on the reading or to do additional research to resolve confusions.

A note on notes

You should take notes during your reading group. I don’t know where you should take them, you might be extremely loyal to your personal Roam/WorkFlowy/Notion/Evernote/whatever, or your group might want to get a shared Google doc going, or something else entirely. Just pick someplace you’ll be able to refer to later.

A motivating anecdote: I was once at a poster session where two people were presenting research that East Bay Biosecurity had read about. I was like, “oh, you’re here, this is great, I know my reading group had a few unresolved questions” and started pulling up my notes on my phone. The researchers were like, “you read our report? And you have notes?!” and that is the fastest I have ever made friends at an academic conference.

A high-energy reading group

I hope these ideas encourage you to start a reading group. These groups can help you learn deeply, build community, and sharpen your ideas.

Right now, reading groups are a great part of my life! I’m currently in a two-person reading group that’s causing me to finally read Seeing Like A State; the book is fascinating and it’s been a great way to connect with an old friend over Zoom. My “read together” group (also two people) is going through papers on longevity; its unstructured discussion leaves a lot of room for looking up unfamiliar terms and methods (e.g. “what is βIII-tubulin stain for?”). At work, six of us are doing an everyone-has-a-role reading group for Gather, since we want to get better at running really effective events. I’m not currently in any academic single-presenter reading groups, but I’ve enjoyed them in the past.

Some other resources on the forum that might be helpful are this reading group guide for EA groups, these Notes on the Art of Gathering, and this AI Governance Reading Group Guide.

Thanks to Aaron Gertler for nudging me to write about reading groups and to both him and Jamie Wahls for providing feedback on this post.

JeremyR4y1

I've never taken part in a reading group (outside of seminars and the like in undergrad), and have no plans to do so, and yet I really enjoyed reading this piece! Thoughtfully and clearly laid out, with novel ideas I hadn't come across before. I'll be sure to pass it on to friends who take part.

I'm glad Aaron nudged you to write this and that he included it in his digest email!

Mau4y7

Thanks for this! I especially appreciated the recommendations for doing 2-person reading groups, and for having presentations include criticisms.

On top of your recommendations, here's a few additional ideas that have worked well for reading groups I've participated in/helped organize, in case others find them useful. (Credit to Stanford's The Precipice and AI Safety reading groups for a bunch of these!)

Break up a large group into small groups of ~3-4 people for discussion
- This avoids large-group discussions, which are often bad (especially over Zoom).
Have readings be copied into Google Docs, with a few bolded lines at the top encouraging people to add a few comments in the doc.
- This prompts people to generate thoughts on the material, and it adds a few interesting ideas to the reading.
Have participants vote on which questions to discuss: digitally share a Google Doc with potential discussion questions, then give people ~5 minutes to write "+1" next to all questions they'd like to discuss.
- Small groups can do this to decide what to have as the focus point of their conversation.
- Alternatively, organizers can use this to break a large group into small groups based on people's interests, like this (adapted for Zoom times):
  - The organizer encourages people to add & vote on questions for ~5 min.
  - The organizer identifies the most popular questions--enough of them that each small group could discuss a different one if they wanted to.
  - The organizer communicates to the group which questions were most popular, and labels each of these questions with a number.
  - The organizer encourages people who are especially interested in some of the questions to indicate this (e.g. by messaging a number to the Zoom chat).
  - The organizer creates groups of 3-4 people, trying to put together people who indicated interest in the same question.
When generating discussion questions, lean away from very vague or big-picture questions.
- Very specific questions (which might be sub-questions of big-picture questions) seem to lead to much more fruitful discussion.

Harry Taussig4y2

I love and appreciate these suggestions! I'll be stealing the idea about copying readings into google docs and am super excited for it.

Tessa A 🔸4y1

These are great ideas! I love all of the practical zoom-call-management suggestions. Splitting into breakout rooms based on upvoted questions in a Google doc sounds quite fun, I may have to try that.

sbowman4y7

Riffing on this, there's an academic format that I've seen work well that doesn't fit too neatly into this rubric:

At each meeting, several people give 15-30m critical summaries of papers, with no expectation that the audience looks at any of the papers beforehand. If the summaries prompt anyone in the audience to express interest or ask good questions, the discussion can continue informally afterward.

This isn't optimized at all for producing new insights during the meeting, but I think it works well in areas (like much of AI) where (i) there's an extremely large/dense literature, (ii) most papers make a single point that can be summarized relatively briefly, and (iii) it's possible to gather a fairly large group of people with very heavily overlapping interests nad vocabulary.

emmabluemke3y8

I agree, this is a common format I've experienced in academia.

For what it's worth, I've found that it sometimes evolves into unnecessary criticisms of the paper, and sometimes the criticisms aren't really correct (i.e. the author isn't there to defend the method and perhaps the presenter hasn't quite understood the paper or reasoning themselves).

I've started to believe that this reading group format might actually contribute to why a lot of PhD students feel so frozen/overwhelmed when writing papers of their own... they watch perfectly fine papers get ritually dunked on once a week, and then those criticisms get embedded into their inner critic and sabotage their writing progress! :-)

Harry Taussig4y5

Thanks for writing this post. :)

I like how you accept that a low-commitment reading group is sometimes the best option.

I think one of the ways reading groups go wrong is when you don't put in the intentional effort or accountability to encourage everyone to actually read, but you still expect them to – even though you're unsurprised when they don't read. But then, because you wish they had read, you still run the discussion as if they're prepared. You get into this awkward situation you talked about where people don't speak since they don't want to blatantly reveal they haven't read.

flowo4y2

Thank you for the write-up! Great idea for assigning different roles instead of just having one person present.

And 100% agree about taking notes and saving them somewhere accessible.

EA Forum Bot Site
EA Forum