Decompiling Oppression #25
Sam McVeety
Let's talk about Timnit Gebru, a phenomenally talented AI researcher, whose contributions to the field will hopefully outlive her employment at Google. I spent a good part of the last week trying to sort out my own feelings about the ordeal, and I found it helpful to return to an idea that we've discussed here in the past: centering the victim perspective (rather than the perpetrator). I won't pretend that this is a simple topic, and want to acknowledge up-front the complexities of a white (cis, male) author writing about the problems with white (cis, male) dominance at the crossroads of industry and academia.
First, I want to speak directly to some of the folks who (like me) view Jeff Dean as something of a giant in the history of Google. He (and Sanjay) are responsible for designing systems without which jobs like mine wouldn't exist. For us especially, we must be extremely diligent not to fall into the trap of framing this issue through a perpetrator perspective. Such a perspective requires that every actor who is part of the systems causing harm meet with our conceptions of what such a perpetrator looks like (and critically, if we can't imagine someone as a "villain" then we are apt to reject the idea that harm is occurring). When we frame things this way, we cast ourselves into a binary that straightjackets our moral thinking. (Also, consider the perspective of someone who views Timnit as a groundbreaking hero for all the many communities she represents.)
Instead of falling into the above trap, we must first and foremost focus on the victim's voice. Timnit is experiencing harm. How do we know? Because she is telling us. Just a few weeks ago, we identified our own frequent inability to recognize injustice in our present. Richard Delgado and Jean Stefancic suggested that one of the ways to remedy this failing is to elevate, amplify, and give power to subordinated voices.
This case is further complicated in that it exists at the intersection of academia and the tech industry. In the latter case, readers are likely familiar with many of the existing inequities and structural oppressions that exist (from here and other sources), but I'd like to provide additional background on the academic context. For that, we turn to Richard Delgado (again) and The Imperial Scholar, a powerful essay on how gatekeeping and structural racism can exist in the legal academy (and by direct analogy, the academic world more broadly) even in the hands of well-meaning majority authors. He describes the strange trend that he observes, where white authors write about civil rights, citing other white authors, and so on, in a cycle. While said authors may have noble aims, he finds that they still miss the mark, because they fail to honestly reckon with the historical roots of injustice -- they "concentrate on the present and the future and overlook the past."
The same applies to conversations around technology and its applications, where majority groups often present a "rising tide lifts all boats" narrative. While it may be more comfortable to focus on the future of technologies like ML/AI, we do a disservice to many marginalized groups by not situating that discussion in a larger context around justice (in this case environmental, social, etc.), and how our pursuit of new advances may in fact exacerbate these historical inequities. We must seriously ask questions like "will training advanced models accelerate climate change?" and "will language models silence and erase smaller communities?", and be willing to say no to some advances if the price is too high (and remember that setting this price itself is a matter of perspective).
When it comes to critiquing majority narratives, Delgado notes the advice that he received (and followed) regarding his own career -- effectively to keep his head down until he was granted tenure, at which point he would have the freedom to discuss "riskier" topics. When this happens, we must ask what is lost? Not only are we silencing voices that are a critical part of the discussion, but tenure is, in some ways, a "best case", because it has a clear timeline and hurdle.
If we turn the same analysis to the tech industry, when is "tenure" achieved? It's not necessarily a certain level, because performance reviews are an ongoing process. Even with the accrual of structural power that a given job title or level might entail, there is still the possibility that one's work will be deemed "objectively" insufficient for their job, perhaps by an individual (or individuals) that might not have the lived experience to understand why someone else might find that work not only sufficient, but deeply urgent. In this way, these urgently needed voices are continuously pushed to moderate, second-guess, and limit themselves, when they should be shouting at the top of their lungs.
It is in this larger context that we situate the deep concerns and anger over the loss (and pervasive absence) of voices like Timnit's. My aim here is not to give you an answer. It is to give you the tools to think more critically about the harm that individuals like Timnit are experiencing.
Here are this week's invitations:
Personal: Reflect on a time when you've been in the position of challenging "conventional wisdom", and how your identity helped or inhibited you from being heard. How might you have perceived this situation differently if the leaders you were challenging were more like you, or less like you?
Communal: Find a group to reflect on some of the critiques above. If it helps, read key passages aloud and then observe some amount of silent reflection afterwards, prior to discussing. Reflect on the emotions that you are feeling. Are you feeling curious? Apprehensive?
Solidarity: Support AddisCoder, a free intensive 4-week summer program in Addis Ababa, Ethiopia introducing high schoolers to programming and algorithms.
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
FAQ
Can I share this newsletter with non-Googlers? Yes! Feel free to forward this note externally; it does not contain confidential information.
Is this an official Google newsletter? Nope. The views expressed in this newsletter are not the official position of Google, and we are not affiliated with any particular ERG.
I am leaving Google or transferring to another bet. Can I still receive this newsletter? Yes! You can join the external list here.