Accidentally in Code

Tag: bias

  • #GHC15 – Brian Nosek: Solving Implicit Bias in Gender and STEM

    #GHC15 – Brian Nosek: Solving Implicit Bias in Gender and STEM

    Credit : Flickr / new 1lluminati
    Credit : Flickr / new 1lluminati

    Start far from implicit bias, at the implicit associations of the mind.

    Our understanding is mediated by our sensory systems. The mediation of these and all cognitive architecture, means that reality and our experience of reality, is not the same thing. Lots of inferences our mind makes to help understand reality.

    This gap. First way is illustrated by the McGurk effect. People hear different things depending on whether you watch him. Sound you hear in your head will be different as a function of whether you are looking at him. This ought to be terrifying. Because we live out lives as though the sounds that occur in our minds are the sounds that happen. But they are our minds best guess. Audio of ba ba, video of gaga, da da is what gets inferred. Reasonable inference – half way between the two.

    Limited consciousness, want to be as efficient as possible, delegate as much as possible to unconscious.

    Picture flashes – horse or frog? Hard to see the other one. As soon as your mind has a first impression. To see the other have to undo the work already done, mind doesn’t want to do that, wants to be efficient. Very hard to go back. Like defaults.

    Image with two shades of grey. In perception care a lot about edges – edges help us know where objects are. Even though know, still see them as different. “Perception is not subject to reason.” What we get to decide is what we do.

    Squares can be experienced differently because of the things around them. Important analogy in social perception. Baby and the jack in the box, asked q’s about video. “What were the emotions?” randomly assigned participants baby as Joan / John. Joan, likely to interpret as fear. John, more likely to say angry.

    Same experience on the outside, same information. Very different internal experiences. Asked if used gender? Said no. Weren’t deliberately using gender. But getting in, because expectations.

    Been looking at what kinds of associations do people have in their memories? One is the Implicit Association Test.

    Word association test. Good / Bad, Female / Male. Then Bad+Female / Good+Male. Then Bad+Male / Good+Female.

    Men and women both show a stronger association with good and women, women about 3x as much. Putting things together should be easier if they go together in our mind.

    Project implicit – people do these tasks, and get individualised feedback. Data on responses. Most men AND women show stronger association with men and career. Effect of experience: women in careers tend to show a weaker association, but doesn’t suddenly reverse. So much association, regardless of whether we believe it or not.

    If we know when this occurs this also occurs, then we can start to predict the future. We construct expectations about what the world is, which translates into beliefs about what the world should be.

    Variety of kinds of biases. Can go online and try it out.

    Some of these associations, have investigated implications on our behaviour. Replace career+family with STEM and arts. Men+Stem / Arts+Female. Both men and women show on average, individuals different.

    Where am I welcome? Where do I belong? These things are inferred by the environment. They aren’t culturally free, they are culturally bounded. Association is effected by what happens to them in their careers.

    “Estimated probability of majoring in Science as a function of sex and gender and implicit gender science stereotype” – paper — graph.

    Suggests that stereotypes are linked to important behavioral outcomes.

    Look at whether linked to behavioural decisions by others. E.g. faculty assessment of resumes.

    Women: judged less competent, less hireable, offered 4K less.

    Contested result. Another recent study found advantage. A lot of debate about under what conditions advantage occurs.

    Looked at data from a number of nations. Different across countries. Correlates with gender difference in science performance. Stronger stereotypes, bigger gap in performance. Suggests these things are culturally linked.

    Centre for Open Science. Mission driven non-profit. Primary tech. All free, OSS. No monetization.

    Mission: improve openness, integrity, and reproducibility of scientific research. Belonging best predictor of whether people stay.

  • So, You Finally Have a Woman on Your Team?

    So, You Finally Have a Woman on Your Team?

    odd one out

    Some unsolicited thoughts (OK, advice) for managers who finally have a woman reporting to them.

    1. Meaningful Projects

    None of us got into the tech industry for the casual misogyny and the rampant sexual harassment, and we don’t go into work excited for the possibility of someone mistaking us for the help. Same as the dudes, we want to build cool stuff.

    Advice I give all the time, and tell myself repeatedly when I’m making a decision that means I can’t do something for the “collective good” – the best thing that any technical woman can do for the plight of technical women is be happy in, and awesome at her job.

    As a manager, the best thing you can do for women on your team is give them something that they will find meaningful, where they can show their awesomeness. This goes double if she’s just had a traumatic experience – remind her why she got into this industry in the first place.

    This is called sponsorship. I’ve noticed white men often struggle with this concept, but they often don’t have a word for it because it’s just something that happens for them. Make it happen for the women on your team, too.

    2. Accept the Possibility of Bias

    All the data shows that women (for the most part) are not treated equally. Take the time to admit that statistically, it’s unlikely that you are without bias. And given those statistics, the probability that everyone around you is without bias is a vanishing impossibility.

    Depressing? Yep. But this is the world we live in.

    I would never advocate giving women on your team more reason to worry, but as an internal consideration this can be helpful to have. If they’ve internalized enough statistics to worry, never deny that worry – the statistics show that it is entirely rational, and in this industry we pride ourselves on rationality, right?

    Things to look for: Ideas being repeated without credit. Judging women on past performance and men on “potential”. Code reviews can be a place for men to exert, or resent (perceived) dominance.

    Overall

    Here’s the thing – women don’t want to be treated differently. These suggestions are not about “special treatment”, they are about the internal work that we can do to ensure that women are, actually, treated equally. Or at least more equally – because truly equally is a long way off.