Review of FHI's Summer Research Fellowship 2020

post by rosehadshar · 2020-11-19T11:32:03.919Z · EA · GW · 2 comments


  Summary of activities
  Lessons learned
  Goals of the fellowship
  Detail on activities
    of activities
    of aspects of the fellowship
      The fellowship as a whole
      Length of the programme
      Program components
      Mental health
  Costs of running the fellowship
    the SRF organisers
  Benefits produced by the fellowship
      Career updates
      Knowledge and skills
      Other effects
  Comparing costs and benefits

This post reviews the Future of Humanity Institute’s Summer Research Fellowship 2020 in detail. If you’re in a hurry, we recommend reading the summary of activities [EA · GW], lessons learned [EA · GW], and comparing costs and benefits [EA · GW] sections for a quicker take.

Thanks to Owen Cotton-Barratt, Max Daniel, Eliana Lorch, Tanya Singh and the summer fellows for reviewing and improving this post.

Summary of activities

Lessons learned

Meta: we’re stating these confidently to make clear which direction the summer fellowship has updated us in. Obviously this was just one summer programme, and in many cases there might be important differences which mean that these takeaways wouldn’t generalise. The takeaways are in rough order of importance according to us.

Goals of the fellowship

The main goal of the fellowship is to help young people to do better things from a longtermist perspective than they would otherwise have done.

We expect most of this change to happen through:

We try to meet this goal by:

Detail on activities

Timeline of activities

Summer fellows

There were 27 fellows in all. Below is a list of those fellows who explicitly gave us permission to name them in this post. See here for more extended bios.

NameBackgroundFocus during the summer
Christopher McDonaldPhysics undergrad, now statistics masters3D chips
Florian DornerMasters in maths, now masters student in science, technology and policyMeasuring Progress in Deep Reinforcement Learning Sample Efficiency
Georgiana GilgallonPhilosophy undergradVarious topics in philosophy and macrostrategy
Gustavs ZilgalvisPhysics undergradVarious topics in technical AI safety
Jenny XiaoIR PhD studentDual moral obligations and international cooperation
Joel ChristophGlobal politics and economics masters studentThe semiconductor talent landscape
Jonas SandbrinkMedical doctor in trainingBiosecurity risks associated with vaccine technology; economic incentives for novel vaccines
Joshua MonradHealth policy, planning and financing masters studentEconomic incentives for novel vaccines
Julia VooPolitical science PhD student, working in tech policyThe relationship between international standards and AI
Kwan Yee NgIR undergrad and AI governance researcher/independent researcher, now masters in China studiesAI crash projects
Megan KinnimentPhysics masters studentFactors that influence how people morally value different beings
Morgan MacInnesPolitical science PhD studentThe relationship between AI, population and military power
Nicholas EmeryPolitical science PhD studentEstimating an AI production function
Nithin SivadasElectrical Engineering PhD student, now a postdoc at NASAThe statistical nature of threats to historical civilizations
Nuño SempereIndependent researcher and forecasterOptimal allocation of philanthropic resources
Siddhanth SharmaJunior doctor, now masters in public healthMetagenomic sequencing and public health surveillance


Evaluation of aspects of the fellowship

The fellowship as a whole

We asked fellows the following questions:

We think these are fairly good responses overall. The minimum responses to both questions came from the same fellow.

Length of the programme

Fellows’ opinions

Mentors’ opinions

Our thoughts

Program components

MentoringAll fellows bar 3 were assigned a research mentor who met with them roughly once a week. 


These seem like the two most important parts of the fellowship structure. In other sections of the survey, multiple fellows stressed how valuable these things were.

See below [EA(p) · GW(p)] for more details on mentoring.

Setting up the 1-1 introductions was time intensive, but seems worth it.

Intros to FHIers Rose facilitated around 80 1-1 introductions between fellows and (primarily) FHI researchers


1-on-1s w/ fellowsEvery week, fellows were paired with one other fellow to have a 1-1


1-on-1s w/ Rose/ElianaRose had at least one 1-1 with every fellow, and checked in more frequently with some. Eliana had a range of 1-1s. The content of these 1-1s was usually checking in, discussing problems and progress and getting to know each other.


This seems valuable but was high time cost. Overall we would do it again, as it helped us to run the programme better and also mitigated some risks.
SocialsWe organised a handful of socials at the beginning of the fellowship, and fellows then organised additional socials themselves.


This was scored relatively highly even though we didn’t put much organising effort into making socials happen. 8 fellows commented that they would have liked more socials, so it’s probable that there was more value to be had here, though we’re unsure if this would have been worth additional time investment on the part of the organisers.
Regular FHI talks and seminarsFellows were invited to most FHI seminars and research events (1-2 per week).


These talks were happening anyway so it was very cheap to invite fellows to them. 
Opening sessionOn the first day of the fellowship we held two 1.5 hour opening sessions to launch the fellowship (2 because time zones meant that it was not possible for all fellows to attend at the same time).


Career planning workshopsA collaborator ran an optional career planning workshop in the 5th week of the fellowship, which ~8 fellows participated in.


Worth running, though interestingly two people rated it as neutral.
Midway sessionAt the end of the third week, we ran two 1.5 hour midway sessions, to check in with people and provide an opportunity for fellows to reflect and reset.


The project review templateIn the 5th week we shared an optional template for fellows to use to review their projects.


This seems worth it as it was such a small time cost. It’s interesting that it was so much more valuable than the project plan template.
SRF talk seriesWe organised around 14 talks during the fellowship, from researchers in the x-risk space. Some were pre-recorded, others were live. 


This took a fair bit of organising, and possibly this isn’t a high enough rating to justify the cost. We think that putting in more effort to make the talk series really great, or not doing it at all, would have been better in our case. We also think that a large part of why this wasn’t more valuable was that we didn’t provide a good enough  format for synchronous viewing of talks, and that had the fellowship been in person, the talks may well have been much more valuable without any other changes. 
The SRF meta docBefore the fellowship, we shared a meta doc with fellows describing the purpose and structure of the fellowship.


Although it’s a relatively low rating, it seems clearly important for fellows to have some doc giving context and explaining the structure.
Final presentationsIn the final week of the fellowship, we organised opt-out final presentations for fellows. These lasted 10 minutes followed by 5 minutes for questions, and were open to all fellows, mentors and FHIers. 23 fellows gave final presentations.


Given the time cost to fellows and organisers, this may not be worth doing in future (or better formats might make it more valuable).
The project planning template In the first week we shared an optional template for fellows to use to plan their projects.


This seems worth it as it was such a small time cost.
Check in group meetingsAll fellows were assigned to a check in group with 2-4 other fellows, to meet once a week to discuss progress and any topics of interest. In the second week, we surveyed fellows on how their check in groups were going, hoping to be able to intervene where things weren’t working, but few fellows responded and we didn’t take any significant actions.


This score was pulled down somewhat by the responses from one group which didn’t go well (the score would be 0.96 excluding that group). 

If this is repeated, it would be good to have a better way of intervening on check in groups when they are not working (e.g. by organisers attending all groups once, or checking in with every fellow individually). 

In spite of the low score, we expect that given the remote nature of the fellowship and the large number of fellows, having smaller groups to talk with regularly may have been important for creating a friendly, supportive culture. At least two check in groups continued meeting after the fellowship, additionally suggesting that this was valuable for some fellows.

Lightning talksIn the second week, we ran three lightning talk sessions. Each fellow had 5 minutes for introducing their work and answering questions. The talks were open to all fellows, mentors and FHIers.


Probably not worth doing, but it’s a bit hard to tell. Fellows gave longer talks than would have been ideal, and tried to cover a lot of information. If we’d been able to better prompt fellows to give really snappy introductions to the motivation for their projects, this might have been worth doing.
Weekly slack check-inEvery Monday all fellows posted what they had done last week, what they were planning to do next week, and any problems or worries in Slack.


This was probably worth it as it was cheap, one of the only ways we created surface area between fellows in different check in groups, and wasn’t rated negatively by anyone.
The discussion group spreadsheetA fellow set up a spreadsheet to coordinate discussions before and during the fellowship.


This didn't really take off, but we think that it was worth it in expectation, and probably worth it overall (as 7 fellows rated it as valuable).
The default online officeIn the first week, we experimented with having a default online office for one day. After the experiment we decided to implement this as a default office throughout the rest of the fellowship.


This seemed costly, as it hampered some more energetic attempts by fellows to coordinate online office spaces. We think that creating a really good online office space could be very valuable, but that cheap versions may be worse than nothing.


For fellows

We asked fellows the following questions:

We think that subject expertise is a nice to have property in a mentor, but not strictly necessary for valuable mentoring pairs to be formed.

For mentors

We asked mentors the following questions:

In one sense, the ratings mentors gave seem quite low. However, they clear the bar of ‘worth my time overall’.


We asked fellows ‘Do you have any comments on the culture of the Summer Research Fellowship?’

We think the culture was surprisingly good for a remote fellowship, and we guess that it helped to increase engagement, support those experiencing difficulties, and deepen fellows’ experiences.

Mental health

We think that given that the fellowship was remote, this rate of mental health difficulties isn’t surprising, and also that the counterfactual is unclear given the global pandemic. 

We also think that anyone having difficulties with mental health that are exacerbated by the fellowship is bad, and we should endeavour to do better. In particular, we think future organisers should think more about ways of mitigating the mental health risks of the fellowship being open ended and (in its current form) having many and therefore unclear paths to success. It’s worth noting that we think that the opposite extreme (of a very clear and particular goal) is likely to be challenging for other fellows, and that the right solution will take variation among fellows into account.


We created a separate Slack workspace for the fellows and their mentors. Mentor engagement with the Slack workspace was optional, and in the end, minimal. We did not add fellows to the regular FHI Slack workspace, as it contains potentially sensitive information, and we didn’t want to flood the Slack with new members.

The Slack worked well as a coordination tool, but didn’t become a rich intellectual space. There weren’t many object-level research discussions, or much social engagement.

Not adding fellows to the FHI Slack worked well. We think we still managed to introduce fellows to FHIers in a way that was valuable.

Costs of running the fellowship


To mentors

To the SRF organisers

To fellows

Benefits produced by the fellowship


To fellows

Career updates


Knowledge and skills


This list contains outputs where there was a publicly accessible link, and/or where fellows were happy to include their name and email address against their work. There were many other outputs where neither of these things applied, which we have excluded from this list.

Note that in some cases only parts of the output were produced during the fellowship. However, we think that for most if not all of these, these texts would not exist, or would have been published later or with worse quality, if not for the fellowship.

Completed outputs:

Notable works in progress include:

Other effects

To mentors

Comparing costs and benefits

We think that the fellowship was pretty clearly net positive, but that it also wasn’t clearly an outstanding success. 

In particular, at the end of the 2019 CEA fellowship, we think there were proportionally more clear success cases to point to than there are at the same point for the 2020 FHI fellowship. This might in part be because there were fewer fellows in 2019, which might mean a) fellows had more input per head and this created more value, and b) organisers had a better sense of how individual fellows had developed, making it easier to spot successes. It might also be because the fellowship was remote and this made it less valuable. Small sample size and frequency of outliers might also mean that this is explained just by luck.

We expect more of the impact of this fellowship to become clear later as fellows progress in their careers, and intend to survey fellows again 6 months after the fellowship and possibly 2 years after.


We’re sure we made other mistakes too, but the ones that stand out in chronological order are:

We may consider sharing some materials and resources with organisers of similar programmes in future. If you think these materials would be useful to you, please email explaining why this would be useful, and we'll consider your request.


Comments sorted by top scores.

comment by MichaelA · 2020-11-21T08:06:21.026Z · EA(p) · GW(p)

Thanks for this post - I found it interesting, appreciate the time you took and detail you provided, and expect the post will be useful for various other organisations who run or consider running similar programs in future.

comment by Max_Daniel · 2021-02-24T14:21:29.059Z · EA(p) · GW(p)

Two Summer Research Fellows - Joshua T. Monrad and Jonas B. Sandbrink - and collaborator Neil G. Cherian have since published a paper they worked on during SRF in Nature: Promoting versatile vaccine development for emerging pandemics.