Toxic Until Proven Healthy

Why people don’t trust *your* startup’s culture

cw: abusive work environments, references to violence

“Innocent until proven guilty” is a great legal theory. No one wants the state to abuse the great power it holds over individual citizens. Making sure that people can’t be fined, imprisoned, or executed unless we’re real sure they did it is a vital check on state power.

Individuals, however, tend not to have the right to fine, imprison, or execute others. (They might have the power to, but unless your name’s George Zimmerman we usually call that a crime.) The ethical rules for individuals who are trying to keep themselves safe in a broken system can and must differ from the rules an ethical state must follow.

For example, Woody Allen was never formally charged with molestation. Regardless of what you think of that: would you trust him alone in a room with your daughter?


The tech industry has historically failed women and other marginalized groups. The modern startup system is one of the vectors of this failure. It reinscribes larger issues and creates issues of its own. Some examples of its failures include:

Proving that many startups are broken is outside the scope of this essay. I’ve provided the links above for those who would like to read more about that. Instead, I’d like to take that as a given and move on to talking about trust: about the fact that many startups being broken renders it harder for women & other marginalized people to trust any given startup; about the ways that these startups harm the larger community by breaking that trust.


schrodinger’s rapist

When a woman meets a man, she knows that he is probably not a rapist. She also knows that he is possibly a rapist — there’s about a 5% chance that he is. One in twenty men isn’t very many, but the consequences of immediately trusting that a man is in the 95% are potentially far worse than the consequences of proceeding with caution. One common metaphor for this caution is that of Schrodinger’s Rapist — unknown men must be treated as a superimposition of friendly likelihood and dangerous potential until they prove themselves one way or the other.

Similarly, women (or otherwise marginalized people) thinking about employment with a tech firm know that many firms are toxic — they don’t even have the statistical comfort of knowing that it’s only one in twenty. (Indeed, anecdote suggests that it’s far more than five percent.) Taking on a job at a firm is a huge commitment. Once onboarded, it’s difficult to leave — one must not only face the logistical difficulties of lining up a new gig on the DL, but also the potential stain on one’s record if one leaves a toxic job “too soon.” Continuing to work in an abusive work environment is likewise dire, and in some cases can induce PTSD.

It’s a small wonder that many, when faced with the risk of that commitment, choose to assume that company cultures are toxic until proven healthy. It’s a small wonder that many choose to leave the industry rather than seek out that proof.


culture smells

Job candidates, especially but not exclusively those from marginalized groups, often actively look for “culture smells” when considering positions, at both the application and interview stages. Often, these culture smells are most apparent to more experienced candidates, whether they’re black or female or just a guy who’s seen “passion” used as a code word for “burnout” one time too many.

Culture smells can include poor language choices in job postings, “benefits” that seem designed to discourage work/life balance, and social signals about office drinking culture, cliquishness, and/or hyper-competitiveness. A lack of existing diversity on a team may also serve as a culture smell. While there are laughably bad job ads out there — one late 2013 job posting had separate, tailored ads for male and female applicants, and the women’s ad cited shoes as a benefit — culture smells are typically more subtle. This does not mean that people are not attuned to them.

(Like “proving” that startup culture is broken, a comprehensive examination of culture smells is somewhat beyond the scope of this article. When I solicited examples on Twitter, I got far too many good responses to include here; I’m going to be writing a followup piece breaking down broad trends, and in the meantime I encourage you to check out @drawingtype’s Storify of the conversation.)

Like code smells, culture smells are an imperfect predictor of underlying rot. It’s possible that workplaces with culture smells are perfectly good workplaces. However, given the alternative, few people will choose to work someplace with culture smells — and many tech workers from marginalized groups consider “leaving the industry” to be a viable alternative.


the limits of signaling

It’s possible, of course, for companies to work to reduce both their apparent cultural flaws and their actual cultural flaws. But how can they convey the latter work to outsiders?

The mere absence of culture smells, after all, just proves they know how to use Joblint. Transparency about business processes is helpful, but absolute transparency about business practices does not always make business sense and absolute transparency about HR matters rarely makes legal sense.

I’d say, hey, maybe do something active!

But, well, GitHub tried that. GitHub started a speaker series about women in tech being awesome. Then Julie Ann Horvath, the GitHub developer who’d founded the series, resigned due to ongoing harassment. It’s now clear to everyone that, despite Passion Projects, GitHub was not a positive place for women in tech. I hope its new HR lead is able to turn things around on that front, but Horvath’s treatment makes it clear that the presence of pro-diversity initiatives does not necessarily indicate a culture that’s safe for diverse individuals.

If positive action isn’t always enough, what is?


Julie Ann Horvath’s harassment at GitHub is just one drop in a very, very large bucket. Every incident — whether it’s public or just whispered about — decreases community trust just that little bit. The plural of anecdote is not data, but the sheer number of anecdotes is overwhelming.

Maybe you’re fine, you, personally. But a lot of people are not fine, and I don’t know which kind you are.


It’s trendy to end essays with a call to action. But there’s no easy action I can recommend, here. Slapping whitewash on structural problems doesn’t mend the underlying joists.

This community, through all of our actions in it, has broken many people’s trust. Trust, once lost, must be earned back slowly.


I am not paid to write. Please consider supporting my work on gittip.


Many thanks to the people mentioned in the “culture smells” Storify for their help with this piece.