I have not felt disadvantaged for being a woman in AI.
Three immediate clarifications
- My experience is likely a consequence of obliviousness, luck, and privilege.
- This is not a counter data point to others. It is simply an additional one. That is, this post should not be used to deny that gender-based discrimination exists in the field or negate troubling experiences other women in AI have shared (many of which I am personally aware of).
- Why am I sharing this? To encourage junior women AI researchers to continue in the field; to tell them that it is not all bad.
I have been struggling for over a year whether or not to share my data point.
The first time I felt the urge to share was when I was coming across women in STEM sharing their experiences. My motivation was to have a more complete distribution of experiences being shared. To say that my data point exists. I exist.
This urge resurfaced whenever I was asked to serve on panels for Women in AI. I struggled with deciding whether or not to serve.
On one hand, I did not think of myself as a woman in AI. My gender was not a part of my professional identity as I perceived it. I imagined that when inviting me, organizers assumed that I must have experienced gender-specific struggles. That I must have perspectives, tips & tricks, inspiring stories to share that would help other women in the audience. But I don’t have those — not gender-specific ones. So perhaps I shouldn’t serve.
On the other hand, there aren’t enough women in AI. Especially at my stage of their careers and beyond. Studies show that seeing people like you a few steps ahead plays a significant role in keeping you on the career path. And so if the women like me who have thrived don’t show up and engage, that would be unfortunate. I felt I would be letting the community down by not serving on these panels. So I served.
But I struggled with being authentic on the panels. With representing the part of the distribution that my data point belongs to. Ironically, I felt like an imposter and a hypocrite.
Was representing the complete distribution important though? Perhaps not. Especially because it suffers from survivorship bias. I figured too many women are being disadvantaged. In fact, too many individuals from different groups are being disadvantaged in different contexts in different parts of the world. It is clear that things as they stand are unacceptable. It is clear that change is needed. And for change to come about, we need to shine the spotlight on parts of distributions that highlight the problems. We need not cover entire distributions.
That settled it. My data point need not be shared.
Until about 6 months ago.
I was part of a team building exercise with a group of colleagues. It involved filling in this template and sharing it with the group.
My name is _______ and I am from _______. One thing you cannot tell just from looking at me is that _______. This is important for me to tell you because _______.
This was a safe space among individuals I trust. I shared:
My name is Devi and I am from India. One thing you cannot tell just from looking at me is that I have not felt disadvantaged for being a woman, or not-white, or short. It is important for me to tell you this because I feel like my data point is a valid one that I’d like to share, but I don’t feel comfortable sharing it more broadly.
A junior female AI researcher in the group asked me why I don’t feel comfortable sharing it. I explained my reasoning to her.
She said that she for one is glad to hear me say this. She thought it would be useful for junior women considering careers in Computer Science or AI to hear this. They may be scared away from the field if they only hear about bad experiences. They might find my data point, and more generally, hearing of experiences and stories from women who haven’t struggled much, reassuring. (I assume I am not all that unique?)
This made a lot of sense to me. And so, with a bit of apprehension, here we are :)
Any reactions, stories, perspectives, feedback, or questions are very welcome.