Women in machine learning

At this year’s Machine Learning Communities Day hosted by Google and Seldon at Campus London, machine learning community leaders in the UK and Ireland gathered to discuss some of the bigger issues we're facing as a community. One of the issues that was often reiterated was the issue of how to open our doors to more women. During a group discussion a few of the members started brainstorming about what the problems we're facing are, why we're seeing them, and how we could resolve them. In the group were two women and four men, a less-than-half-but-better-than-average representation of women.

Note: The points below are based on the impressions and opinions (in case of the women) of those present. No research was undertaken, and we don't claim that this is neccessarily an accurate or exhaustive list. The word 'meetup' is used throughout, as most of the people present were organisers of meetup groups, but the concepts apply to all events-based communities.

After brainstorming we discovered the following issues we were facing:

1. The format of the meetup: Do you start off with the social aspect to then move on to the technical bits? An extended social part of a meetup event, in particular if beer is involved, tends to lead to unwanted flirting.
2. The topic of the meetup: In the machine learning community in particular, the topics discussed tend to lean towards boy-ish subjects. E.g. how to beat professional poker players using ML or more highly technical talks aimed towards a male-saturated programmer culture, in particular compared to the data-science community.
3. The atmosphere of the meetup: If the event is held at your local pub your women membership may be reluctant to go. Additionally, the way organisers and other members greet and welcome, in particular first-time goers, can make or break the impression of your community.
4. The key people in the meetup: Are talks or presentations given by female speakers? Do you have a female presence in your leadership team? It's hard to be the first, or the odd-one-out of any community, and so women may feel uncomfortable joining a group or event where they'd be the only woman.
5. The purpose and outcome of the meetup: Women tend to want a clear purpose for going to the meetup and expect something to take away from the event. An event with the title "Let's meetup to talk about sentiment analysis" may not be as interesting as "Learn how to build a sentiment analysis model in 30 minutes".
6. The refreshments at the meetup: Many of us have to skip our dinners to go to meetups, and the pizza-and-beer expectation that many of us hold when going to meetups (or is that just the ML ones?) can actually be a bad thing.

In answer to each of the issues we came up with the following possible solutions:

1. We firmly believed that networking and socialising is almost a requirement for a successful meetup event. So a solution to this problem would be to hold the social bit after the event. This way anyone who feel uncomfortable socialising, regardless of gender or reason, will feel free to leave at any point.
2. Machine learning is a widely-applicable subject. There's no reason why there can't be talks on more gender-agnostic or topics that have a wider female audience. For example, the women present mentioned Creative AI (no doubt inspired by the presence of @elluba at the event, although not in the group). Further, topics such as the ethics and social impacts of AI and ML could widen the demographic interested in the subject to a more non-technical audience. An event co-hosted with a law-oriented community was cited as having a gender-balanced attendance.
3. To solve the issue of the atmosphere there's no need to rent out a fancy ballroom for an evening. Does your company allow you to host events in one of your larger meeting rooms? Perhaps there's a local community space such as The Guild Co-working hub in Bath? A clean, bright venue appears to be a decision factor. Further, a nice welcoming of any new members never goes amiss. If these newcomers are attending an event where the technical aspect may be daunting, pair them up with someone you know are skilled and patient enough to help them out.
4. If you have a large meetup group, chances are you could use some help with the organisational aspects of your events. Consider asking one of your female members if they'd like to join your leadership team. The data science community appears to be more diverse than the ML-community. There's plenty of experts and speakers to whom you could reach out to give a talk.
5. To sell your events to women define clear outcomes of your events. An event with the title "Let's meet up and talk about sentiment analysis" may not be as interesting as "Learn how to build a sentiment analysis model in 30 minutes". Hands-on/Workshop events appears to be popular.
6. I cannot seem to remember the alternative to pizza and beer that was suggested, but there's no reason why fruits and soft-drinks can be served, if not instead of, at least alongside the standard beer and pizza.

Can you think of any other problems and issues that women in our community face? Have we misunderstood something, or drawn a wrong conclusion? Please let us know and share your ideas in the comments below.

ML Love,
Roger (@kinbiko)