@samabbott and I have been discussing an idea for a meeting whilst in transit from SA → UK (so may need to come back and edit later). We are very keen to hear what others think about this, what the alternatives might be, and if anyone wants to be involved. We’d like to discuss at the next community call, or organize a separate call specific to this if that works better for people.
The idea
Gather infectious disease researchers that have software development as a key element of their work, for the purposes of addressing some of the “meta” issues. E.g.:
- how do we actually write software? how do we want to be writing software?
- how should we be training people?
- what incentives are good / bad for producing research software?
- what is good research software?
- what would the benefits of more / better research software be?
Depending on interest we could tighten the scope to infectious disease modellers or widen the scope.
Outcomes
The symposium would ideally yield some publications related to the answers to those questions, along the lines of:
- A best vs actual practices working paper (for practioners)
- A case study working paper (for practioners)
- A commentary outlining challenges / way forward (for funders / institutions)
The meeting
To achieve useful outcomes we think we should aim for a week long residence with between 30-50 participants either in the UK or in the US depending on interest from participants and funding organisations.
For the agenda, we thought something like:
- Every attendee presents informally on their tool for ~15 minutes; this would not be a results talk, but more like a developer documentation walkthrough. We’d have pre-organized asynchronous feedback channel, with some prompts.
- Panel discussions/invited talks on key software developent topics from experts outside the infectious disease field.
- Focus group sessions drafting on the working topics - with some scheme for moving people around the groups, to ensure everyone contributes to all the topics.
- Periodic whole group discussion on the focus topics, with at least a closeout summary led by the emergent point people for the focus groups.
- Some optional activities (e.g. pair coding, code review) where we organize matching.
- As a separate-but-linked day on the end of the symposium, organized team meetings or hack-a-thons for particular open source projects (for example epinowcast). This would depend on intereest from projects as to its exact shape/existence.
Funding
Some potential funders in the UK could be:
- Turing
- Software Sustainability Institute
- Wellcome
- Royal Society?
- data.org (@hgruson interested if you have thoughts either on form on or funding)
If US based some funders could be:
- CFA CDC. For the groups applying for the bolus of CFA money, this might be a concrete plan to execute on vaguely outlined training / development program objective(s).
- Gates
Organisers
We would need to find people interested in helping shape and organise something like this. Who exactly this is depends a bit on the location obviously.
1 Like
Mentioned this on CMMID slack, and was reminded about the CMMID workshop funding opportunities. Might be possible to assemble from that (targetted at 1-2 days) + other sources, and leverage some LSHTM administration.
Something I think we want to focus on here is being developer-led vs organised by others as this way we can focus in on content that will be most useful for developers for which there seems to be a bit of a gap at the moment.
I’m happy to find time to chat directly about this. This is a topic clearly aligned with my interests.
My main question at this stage: what would be the difference with existing RSE conferences/workshops? The topics you mention seem to be the topics usually covered at these events.
Do you mostly envision having a different audience who does not necessarily see themselves as part of the RSE community (i.e., probably more traditional research roles?)? Do you expect that the focus on infectious disease modelling and modellers is going to lead to discussions different from topic-agnostic events? Do you have specific elements that are not usually covered in RSE events that you have in mind for this event and that I missed from the first comment?
1 Like
@hgruson I see that @samabbott and I left out the bit about the audience.
Sam should chime in as well, but I’d say roughly:
- for more traditional research positions (though probably role-wise, practically doing something that looks RSE-ish)
- more seasoned scientific programmers
- working at relatively large scale
Essentially, folks that have weathered the pre-RSE era, and accrued relatively large, sophisticated code bases. In the way that most computational scientists are self-taught programmers, some manage to become self-taught developers as well - we’re looking for folks that have gone in that direction.
Thinking generally infectious disease epi computation as a way to focus advertisement / recruitment, modulating a narrower (e.g. ID modelling only) or wider focus (e.g. also bioinformatics-y) as appropriate.
This sounds like a great idea. I’m happy to participate in a planning meeting as well.
1 Like
envision having a different audience who does not necessarily see themselves as part of the RSE community (i.e., probably more traditional research roles?)?
I think you are right @hgruson that a lot of this is covered at specialist RSE events so yes aiming for a more varied audience would be the plan.
Do you expect that the focus on infectious disease modelling and modellers is going to lead to discussions different from topic-agnostic events
Yes, I think this is likely. There are obviously aspects of what we do (like Bayesian model fitting) that are covered less at these events at the blend of these topics (i.e. our kinds of data etc) are quite domain specific. There is also the more general point about within-community action being different to across community action (and there being a role for both of course).
I agree with some of what @pearsonca is saying though I don’t think we want only those lonely 
s
Thinking generally infectious disease epi computation as a way to focus advertisement / recruitment, modulating a narrower (e.g. ID modelling only) or wider focus (e.g. also bioinformatics-y) as appropriate.
Also agree with this.
@dajmcdon that would be great! I think you are a good example of the kind of attendee as well that might not (correct me if I am wrong) be going to specialist RSE stuff so much?
@pearsonca any thoughts on funding for this or more generally how to take it forward?
Waking this back up after a related discussion at @kejohnson9 epinowcast meeting yesterday (Community Seminar 2024-08-07 - Kaitlyn Johnson - Wastewater modeling to forecast hospital admissions in the US: Challenges and opportunities - #2 by adrianlison).
I still think this or something more focussed would be a really useful thing to do. Shall we set up a call to take it forward?
1 Like
I think if this could focus on prioritizing and strategizing about how to achieve some of those common interoperability problems @samabbott outlined in the linked discussion, this could be really useful and immediately impactful.
Also think it would be a great place for early career researchers in infectious disease modeling looking to break into this space, and wanting to understand where the field is at and where it’s headed.
Think setting up a call to discuss would be great.
I also wonder how much overlap the R(t) collabathon will have with these topics. Strikes me as a bit more problem specific rather than “developers in ID modeling specific”, but I see some potential for strumming up interest and kicking off some of these conversations there at least informally.
I’m game for a call. @samabbott and I had a chat this weekend about what are loosely called model-builders, and are kind of trendy in the field at the moment.
Roughly, a model builder takes a model short hand and implements the corresponding, for example, ODE. But people also tend to conflate model builder with a variety of other automations that would be very beneficial to our field - construction of fitting objective function, assembly into pipeline (for fitting => forecast/scenario projection), and so on.
odin/dust is an example of this.
How to lead development efforts for that kind of tooling (which sort of exists in other fields, but for which we have distinct needs / practices / training / etc) would be a great motivating / focusing topic to hang other questions off at an event like this. I work primarily in fully-mechanistic, scenario-based analysis, but I imagine this kind of issue also applies to the more semi-mechanistic, forecast-oriented folks as well.