Reflection by Grace Nho (scroll below).
By Grace Nho
Cornell Tech
On April 25th, David Gray Widder presented a thought provoking lecture on Epistemic Power in AI Ethics Labor: Legitimizing Located Complaints.
When people ask me why I pursue a career in technology, ironically enough, I always answer with, because I am pessimistic about technology. Before I was introduced to the harms of technology, I used to view engineering and computer science as a gateway to better opportunities for people who would not generally have access to them. But learning about the problems, the biases, the discrimination, the exclusivity that has seeped into our technology, perhaps not intentionally, I realize that there is a lot more work in order to undo some of the harm that has already been inflicted.
Ethics is so important, yet still unprioritized to the extent where sometimes it feels de-incentivized. During the talk, David highlighted some of the unfortunate events that have happened in the industry - of which one has not left my mind : “Microsoft axes ethics team - moves remaining employees within product teams”. He mentioned that employees who were initially a part of this team find it harder to achieve goals now that the team has been disbanded and they are not a collective movement anymore. This has led me to a newfound appreciation for collective teams that move to better the industry that we are in. It’s so hard to be vocal about certain issues when you move alone, but when we start to think about what we can do to make technology more inclusive together, we start to empathize and become more aware of things that could sometimes be overlooked.
David raises a good question about epistemic power - how can we share this epistemic power with others that generally might not? Before we talk about how we can share epistemic power, we have to understand what epistemic power is. Epistemic power refers to the ability or authority to produce, shape, or control knowledge and understanding within a given context or field. In the context of the lecture, it more specifically talked about how epistemic power in AI is based on automation and quantification to achieve legitimacy and how it can be the cause of the imbalance because this approach delegitimizes located complaints and diminishes the importance of embodied and lived experiences.
In the paper, David proposed the concept of "humble technical practices," which are quantified or technical approaches that explicitly acknowledge their epistemic limitations. By doing so, these practices aim to flatten hierarchies of epistemic power and make space for a more inclusive and equitable understanding of AI ethics that incorporates diverse perspectives, including lived experiences. While I agree that that is a good way to approach the problem, I began to think about how we can incorporate this ”humble technical practices” into rather generic software engineering roles.
Which brings me back to why I want to be a part of the tech space. These power imbalances are tricky, they require the epistemically powerful (generally us) to cede and share the power. There was a deep discussion during the seminar about if and how we can share epistemic power and several solutions and suggestions were shared. This included participatory AI approaches, practice talking from your own lived experience, seeking to elevate those with lived experience. While the solutions may sound obvious, you come to realize that it can actually be idealistic, rather than pragmatic.
Ethics in the field of technology is a difficult space, but much needed nonetheless, to navigate. It’s even more difficult to come to a consensus because diverse lived experiences will have different takes on ethics. This seminar and talk was thought provoking and brought me back to my purpose of wanting to pursue a career in the tech space. How will I put “humble technical practice” into actual practice? How can I try to focus more on lived experiences? That’s for me to find out on the next leg of my career, but it’s a question that I will keep on my mind.