Getting Together: Science Speed Dating

Ok, so this is probably the nerdiest thing I’ve ever seen anyone do when planning a conference, and after nearly seven years at the University of Cambridge, that’s really saying something.

So, along with the free-for-all Cell Polarity conference (see previous post) at the Royal Society that the lab co-organised, we were also invited to organise a much smaller discussion session meeting at Chicheley Hall, a beautiful country house owned by the Royal Society. The idea behind this meeting was to strengthen the research field of cell polarity by allowing some of the main players to get together in a more ‘intimate’ environment to share knowledge and ideas.

This is all great, but how do you go about encouraging people to meet new people and not to just catch up with collaborators? How do you help your scientists make useful new connections and learn about potentially helpful new techniques? And, how do you do all of this in just two days?

Speed Dating! Obviously.

Now, bearing in mind that this conference is being organised by a collection of scientists with an interest in using computers to solve large scale biological problems, that your group of scientists don’t easily divide into two equal groups and that your organisers are all pretty much a massive collection of bio-computing nerds. It would, of course, be far too easy to follow a standard speed dating paradigm. It’s much more fun to collect data on all of your conference delegates and use scientific clustering methods and network building algorithms to pair up the scientists in attendance based on their keenest interests. So, with much credit in particular to Federico Vaggi (a member of the Csikász-Nagy lab) and his prodigious last minute programming skills, that is what they did.

So first you collect your data. Who already knows who? Who works with what (‘known methods’, see figure 1)? And, who would like to work with what in the future (‘wanted methods’)? Then you put all of that together into a computer and group people together based on similar knowledge bases (see figure 2).

You can also build a network of which of your participants have worked together in the past (see figure 3). Each point represents a person and the connecting lines show that they already have an academic connection.

Once you have all your data and a pretty way to look at it you need a set of rules for pairing people up for their dates (This is basically the same thing that happens on most dating websites when they suggest matches for you). At the conference there were two sets of rules used to create ten dates for each person.

For the first five dates, it was about meeting people you’ve never met before and explanding academic horizons. So you can’t pair up people that have collaborated and ideally you want to match people that are on opposite sides of the collaboration network (i.e. haven’t even worked with people who have worked with their match before). People were also paired up based on the things that they knew, meaning that partners had little or no common academic knowledge and so would have lots to tell each other about.

The next five dates were harder. Delegates had said what they know and what they wanted to know, now was the time to match these up. The idea being that partners could talk to each other about skills and techniques that they were interested in learning about for their future work and that may build useful collaborations. This is really difficult because although it’s very easy for you to learn something from your partner it is very difficult to also ensure that they want to learn something from you. Again, you also need to avoid pairing up people that have already worked together in the past.

Even for the small number of people at this conference (about 40) the number of possible pairings is astronomical (something like – 319 830 986 772 877 770 815 625) so even computationally this is a really difficult problem to solve. Through complicated maths you can put together a graph of the favourability of different pairings and solve it to give the best combination of dates.

Apparently it was a great success. Many new conversations were started at the conference and it really helped open everyone up and get them involved in the presentations and discussions even for sessions that they wouldn’t normally be interested in.

There you have it. This is what happens when science gets applied to conference organising; A bunch of crazy diagrams and enforced socialising, something that should definitely be tried more often. As I’ve shown though, it gets really complicated really fast if you start inviting a lot of people, so keep it to the smaller events. Also, don’t come up with the idea at the last minute, it makes life much easier for the poor programmers.

(Image of Chicheley hall sourced from the Royal Society website, all other images of speed dating cluster analysis are copyright Federico Vaggi and must not be reproduced without permission from the author).