Meet Emma Grua: Turning Philosophical Rigor into Policy Impact
Emma Grua, a graduating senior double majoring in Politics, Philosophy, & Economics and Business Administration (Finance), has spent her time at Northeastern applying the tools of philosophical analysis to one of the most pressing legal questions of our time: how should the law regulate artificial intelligence? As a member of the AI and Data Ethics group (AIDE), a highly interdisciplinary research team known for approaching technology policy with a philosopher’s eye for language and argument, Emma has helped shape real legislative conversations at the state level.
How did you become involved in giving feedback to the Colorado State Attorney General’s office about the state’s AI legislation?
During my sophomore year of college, I joined a working group called the AI and Data Ethics group, or AIDE. A year or so into my membership, our group was asked to provide commentary on what was then Senate Bill 24-205 by the Colorado Attorney General’s Office. I guess the Office heard about our work through one of our faculty leads (shout out Professor John Basl!) and thought we might have some insight. The bill was concerned with setting rules and guidelines for artificial intelligence usage, and was particularly focused on its potential for discrimination. We ultimately produced a paper outlining our overall evaluation, general areas of concern, and then several specific issues we wished to flag.
AIDE is an interesting group in that, like much of Northeastern, it is highly interdisciplinary: there are people with all sorts of majors, backgrounds, and interests who come together to do research, engage in good-faith discussion, and ultimately create interesting projects like the legislation feedback. At the same time, it is a fundamentally philosophical group. So, we approached it as philosophy students: we spent a lot of time going over specific words and definitions in the bill, and an equal amount of time reading and discussing outside sources—especially philosophical papers.
You also participated in a project to provide feedback to the Massachusetts House this past semester about a bill regarding digital duplicates. What was that like?
This bill is really interesting (if you, like me, find legislation interesting) in that it has a much narrower focus than the Colorado bill. It is extremely concise, and deals with one thing only: what contract work can legally be fulfilled by a digital duplicate? The Colorado bill was a lot wider in aim and scope, so having a bill this targeted allowed us to get really detailed with our feedback.
What is a digital duplicate? The answer is that even the legislation doesn’t really seem to know. What does it mean to be “highly realistic?” The legislation limits digital replicas to “sound recording, image, audiovisual work, or transmission in which the actual individual either did not actually perform or appear, or the actual individual did perform or appear, but the fundamental character of the performance or appearance has been materially altered.” What is the fundamental character of a performance? What is a material alteration? And even beyond these questions (as one might defend the vague language utilized in the bill as leaving space for reasonable legal interpretation), what about text messages or emails sent from an AI mimicking someone else? Is that a performance or appearance? If you think you are talking to your therapist via text or DM, but you are instead you are talking to Therapist-bot (sidenote: there are a staggering number of people/researchers who are creating them-bots and documenting their findings, which is an interesting though also deeply existential genre of literature), could that constitute a breach of contract?
Also, as the bill deals with contracts, it primarily focuses on the responsibilities that the two parties that enter into the contract have with regard to one another. But what about third-parties? If an artist signs a contract with a venue, and the venue sells tickets to the public, what is the public owed? Do they have the same rights as the venue? It seems, under the wording of this bill, that they do not. At minimum, it seems they ought to be owed some level of transparency.
In writing our feedback we heavily referenced Danaher and Nyholm’s “The ethics of personalised digital duplicates,” which proposes some more concrete principles about what uses of digital duplicates might be considered permissible. Digital duplicates are such a fascinating subject to me, and many others (or so I assume, because I am not sure if the bare handful of people I talk to on a regular basis can be accurately described as “many”) both because of the sort of nebulous and existential questions their existence raises, and the need to provide at least some of these questions with pragmatic (and sometimes regulation-friendly answers).
How was your work and research used by Colorado and Massachusetts?
As I’ve already mentioned Colorado, I’ll focus on Massachusetts. The bill is currently ordered to a third reading, which I’m told is a good sign. I believe this is the final “vote” in the House on the bill, which will then (if all goes well) pass to the Senate. The process is a long one, and I wish I had more interesting updates. Apparently, our state legal system is more concerned with procedure than it is with allowing me an exciting interview.
During your research on AI legislation, what surprised you about AI?
There is actually quite a wealth of thoughtful and well-written information on AI. As a philosophy student, I am biased, but I would encourage anyone who is seeking nuanced assessments of AI to look for philosophy papers on the subject (philpapers.org is great research, and so is Northeastern’s Scholar OneSearch). It sometimes feels as though at every corner, someone is extolling the virtues of AI for some frankly dystopian use-case, and so it is nice for me to know that there is a lot of well-founded and interesting work about it, and our relationship with it, as well.
What surprised you about the existing laws around AI?
I would not say much about AI itself surprised me. I find AI infinitely less surprising than people. And so I think the more interesting and surprising questions all have to do with people. For example, why do we feel such a need to anthropomorphize AI? And what are the harms or benefits of doing so? In AIDE, we tend not even to refer to it as AI as often as we call it LLMs (Large Language Models), perhaps partially out of some rejection of the idea that machines have what might be called “intelligence.” I believe that this anthropomorphization is reflected in the ways that legislatives propose AI regulations: they often feel as though they are trying to regulate a human, who might behave in certain ways or be limited to certain forms of expression, when in fact this is simply not true. I found the paper “ChatGPT is bullshit” to be particularly helpful in exposing my own tendencies to anthropomorphize LLMs.
How did your coursework prepare you for these projects?
I found my philosophy classes in particular to be extremely helpful. I realize that, to other majors, philosophy is a somewhat opaque subject (“what do you do, just think?”). But I think, ultimately, that the skills of philosophy is invaluable. Being able to deconstruct and reconstruct an argument, being able to articulate your own position and interrogate its soundness and validity, and being able to distinguish between normative and descriptive arguments and evaluate each on their own merits are irreplaceable.
My first year at AIDE, we ran a mini-experiment of sorts, to see how well AI would do at philosophy assignments. It ultimately underperformed, as it is simply incapable of the kind of rigorous thought and precise articulation (and also consistent citations) that philosophy demands. I have little faith that it will ever perform substantially better, simply by the nature of its design. You cannot arrive at a well-structured and rigorous argument through imitation, no matter how many rivers your data centers drain of water. Such arguments need to be constructed carefully from the ground up. Thus, if you wish to learn a skill that is both incredibly useful and also irreplaceable by AI, I would suggest trying philosophy. Plus, the department usually has free food.
What drew you to these projects?
Law is a notoriously slow-moving apparatus, and technology often moves at breakneck speeds (sometimes literally). Because of this, there is often a gap between a technology being brought to market and the legislation that regulates such technology. We saw it recently with social media, and we are seeing it again with AI. Commenting on this legislation is both a good window into the regulatory process, and also a good chance for lawmakers to receive feedback from more diverse perspectives.
For those who are interested in philosophy at Northeastern, but somewhat intimidated by the prospect of reading lots of old dead people’s work, I would suggest taking the Technology and Human Values class. One of the professors (shoutout Professor Vance Ricks!) is also a faculty lead on AIDE, and I feel that a lot of the material in the class overlaps with the work that AIDE does, if any of that happens to sound interesting to you.
Did your research change how you use AI in your own life?
No, because I am something of an AI luddite, although it is becoming ever-harder to avoid.
What haven’t I asked you about that you want to bring up?
At the risk of being too much of a philosophy student, and also of sounding preachy, I would like to encourage students to reflect on what they want to get out of college, what they want to get out of life, and how their relationship with AI might foster or impede either of those goals.
What is it that you’re using AI for? Is using it in that way robbing you of valuable knowledge or skills that you might develop by doing it the old-fashioned way (i.e., with natural intelligence)? Of course, we are in school to get degrees, but we are also in school to better ourselves (or at least I hope this is the case). By having AI summarize that reading for you, do you lose the opportunity to sharpen critical analytical skills? By having AI write that draft for you, do you lose the opportunity to learn articulation and perseverance? What kind of person do you want to leave college as: the same as before (but with a degree), or someone who is better, smarter, and stronger in spirit than when you began? Having recently graduated myself, I can attest that I value my growth of character over the past four years above all else I’ve acquired in undergrad, even my now-freakish Microsoft Excel abilities (which I’m sure an AI could match, if not exceed).
Aristotle (cliché, I know) wrote in Nicomachean Ethics that to live a good life is to become a virtuous person. And the pursuit of virtue requires practice. To become a courageous person is to practice acting with courage a million times over. And so, if we rely so heavily on a tool that robs us of practice, and instead offers shortcuts and promises little labor, it is worth asking: do we also rob ourselves of living good lives?
