Local Group Event Idea: EA Community Talks

By Neel Nanda @ 2020-12-20T17:12 (+26)

Premise

I’m involved in running EA Cambridge, and last term I started an event series called EA Community Talks. Group members picked an EA-relevant topic they wanted to learn more about, and prepared and gave a short, informal talk on it to other people in the community. I think this was pretty successful and very low effort to organise, and I’d be excited to see other local groups trying a similar event! My goal in this post is to give a retrospective on how it went and why I think it was high-impact, and advice for running something similar yourself.

My premise behind this was that learning and thinking more deeply about EA topics is super valuable, and that teaching ideas to someone else is a great way to learn, because it forces you to structure and process the ideas. And that deciding to do something like this is effortful and takes agency and thus rarely happens, so it’s valuable to have a public commitment of device of ‘I am giving a talk in X weeks’. 

More generally, I think an important path to impact for a local group is encouraging their more engaged members to be more agenty, explore ideas they’re curious about, and especially to think about career-relevant cruxes and confusions, eg helping them work out their views on cause prioritisation. For many people, doing this is high-effort, and not the default. So a valuable role for a community is to provide structure to make this kind of thing more normal and easier to decide to do. 

I think there’s also value to the people attending the talks, since they get to learn about a new area or idea. In practice, feedback suggests attendees found the talks useful, though I am less excited about this than the impact on the speaker. 

Note that a key part of this premise is that people give a talk on something they don’t already know a lot about, or at least take a new angle on something they’re familiar with. I think much of the value comes from the speaker learning something new, rather than, eg, just sharing their existing expertise.

Impact

In practice, I think this went really well! The interest was fairly high to both give talks (6 talks were given and 5 more are planned so far) and to listen to them (we averaged about 8 audience members). I think people picked interesting topics, and seemed to get counterfactual value out of preparing a talk. (I gauged impact by messaging speakers and attendees afterwards and asking for feedback)

The topics people spoke about:

• Should we prioritise suffering or flourishing?
• A review of Part 1 of the Precipice (ie, is the longterm future really that important?)
• My personal cruxes for AI Safety (ie, how compelling is the case for AI as an x-risk, and is there useful technical work to be done?)
• An attempt to quantify the cost-effectiveness of working on x-risk
• Quadratic voting
• How did the energy sources/economic incentives of historical civilisations shape their values?

5 of the 6 speakers thought they put significantly more effort into learning about the topic as a result of giving a talk (one speaker planned on writing a blog post about the topic anyway, so they think the effort was not counterfactual). Everyone reported understanding the topic better as a result of having to think about explaining it, and generally seemed to have a good experience. I think for at least two speakers (and plausibly more), the talk topic was relevant to their career plans and long-term impact.

My personal experience supports this data: I gave a talk on ‘my personal cruxes for AI Safety’, trying to summarise the arguments for AI Safety I found most compelling, convert them into cruxes, and figure out whether this was actually a career path I wanted to pursue. Personally, I found this very useful, I’m confident I wouldn’t have put in the effort counterfactually, and I feel like I now have much crisper thoughts on which areas of Safety I think are important. And as a result of the talk, I’ve had several conversations about Safety that have exposed me to new arguments. (Though obviously this is biased data, because I was the one who thought this event was a good idea in the first place!)

I also think this is a useful structure to have as a group organiser. I had several conversations with people where they mentioned a topic or question they were confused about, and I talked them into giving a talk about it. I don’t think there would have been a similarly effective thing to suggest they do without this structure. 

I tried to keep the talks feeling low stakes, informal and non-intimidating, so tried to set low standards for speakers, but I was pleasantly surprised by the quality of the talks people gave and the amount of effort put in. Attendee feedback was good, people generally seem to find it useful and interesting, and felt like they learned things. 

I think there were also a bunch of other smaller benefits, eg speakers getting practice at giving talks, hopefully being more likely to do things like this in the future, creating more events for the community, etc.

Doing this yourself

Overall, this was a pretty low-effort event to run, and seemed to work well remotely. The main important thing was to set a regular time for a call, let people know about it, set-up a Zoom call, and convince people to sign up to give talks. 

Most of the effort went into being somewhat pushy with convincing people to sign-up, which I think was valuable. Strategies like posting the idea in a group chat were fairly ineffective at getting sign-ups. My approach was to PM people the idea, or to bring it up in conversation, which both worked fairly well. Several people were interested but reluctant, due to eg not thinking they could give a good enough talk, or struggling to think of a topic, and ultimately had a good experience giving a talk. On the flipside, it’s obviously bad to be too pushy, and some people don’t like giving talks, respond badly to commitment devices or just don’t have the time - it’s a judgement call. 

I think it was important to emphasise the idea that people pick a topic they want to learn about, rather than one they already know about, I missed this nuance when first advertising.

Some prompts I gave people struggling to think of topic ideas:

• What are your biggest confusions for your career plans? Could you learn anything to help with that?
• Are there common ideas in EA that you disagree with, or don’t understand?
• Any topics you’re curious about, but never got round to digging into?
• Is there a possible Cause X you could dig into?
• Set a 5 minute timer, and list as many crazy talk ideas as you can. At the end, see if any are actually worth doing

The exact format I used: Have a 1.5 hour Zoom call every other Sunday evening, have two talks per event, split people into 3-6 person breakout rooms to discuss after each speaker and switching after 45 minutes. I think this worked pretty well, though I doubt the exact details matter. 

Some tips:

• I asked speakers to send me an abstract a week before, which I think was a useful reminder/prompt to start preparing
• Sunday evening was a good time, it meant people had the weekend to prepare if they left it to the last minute
• I told people to prepare a ~10 minute talk, but didn’t care much about timekeeping. In practice, the median talk was 20-25 minutes. It might work better to be stricter about timekeeping, though I expect this to make the talks feel higher effort
• I tried fairly hard to frame the talks as a casual, low effort thing, told people not to stress about preparing, only invited engaged community members (rather than making it a public event), etc. I think this worked fairly well, and made people more likely to sign-up

Conclusion

Overall, I’m very happy with how this went, and I’d be excited to see other groups trying something similar! I think the important part is to have a structure where people can commit to exploring an idea and explaining it to others, so please take these details as just a starting point! Eg, I can imagine “people write an EA Forum post/blog post” being similarly useful. EA Stanford have implemented this idea by having someone give a short talk at the start of each committee meeting, which seems to also work well. 

If anyone plans on trying this out, I’d be very happy to chat more, and I’d love to hear how it goes! You can reach me at neelnanda27@gmail.com