#  Video: Humanity Meets AI Symposium: Building Ethical AI 

 





April 18, 2025

 

 

     ![Humanity Meets AI Symposium](/sites/g/files/omnuum8216/files/styles/hwp_16_9__480x270/public/2025-03/Humanity%20Meets%20AI%20Symposium_0.png?h=087dead0&itok=p5DRo4H3) 

 



 

How might AI influence our understanding of humanity, morality, and meaning-making? How will religious traditions and communities adapt to or shape the ethical frameworks guiding AI development? In what ways can religious perspectives contribute to the creation of a more equitable society amid the disruptions AI will bring to labor, governance, and social structures?

Religion and Public Life hosted a symposium that explored the profound ways in which artificial intelligence is reshaping human society, with a particular emphasis on the role of religion, the transformation of societal structures and capitalism, and strategies to reduce inequities as society responds to the sweeping changes brought about by AI.

The symposium equipped our audience with the tools and frameworks to critically engage with the ethical, spiritual, and cultural dimensions of this transformation. In an age increasingly influenced by AI, this symposium helped identify practical opportunities to shape a more humane and just society.

Embed



 



FULL TRANSCRIPT

\[MUSIC PLAYING\]

SPEAKER 1: Harvard Divinity School.

SPEAKER 2: Building ethical AI. February 28, 2025.

SPEAKER 3: And our third panel is on building ethical AI. And we will start with four wonderful panelists. Jenn Louie is a tech policy and governance expert with leadership roles at Google, Facebook, and Meta. She serves on the board of equal access international and is developing spiritualcareproject.com, a platform for inclusive spiritual care.

She is currently at Harvard Divinity School studying-- no. She's no longer at Harvard Divinity School. She was last year when she was a MRPL here as well. Richard Jereson is CEO and chairman of GBCI. With extensive leadership experience across fortune 500 companies, AI startups, and global advisory boards.

He previously held executive roles at IBM, Nokia, and Toshiba, and has led AI software firms to major industry breakthroughs. He holds a doctorate from Oxford University and has served on the Harvard faculty. James Hughes is an executive director of the Institute for Ethics and Emerging Technologies, a bioethicist, and a sociologist. He is also the associate provost at UMass Boston and has written extensively on transhumanism, AI ethics, and the future of AI of human enhancement.

And John Richards is a senior executive in education, technology and media, specializing in business development and digital learning. He is the founder of CSforEd, an education consulting firm, and has held leadership roles at Turner Learning and the Jason Foundation. His research focuses on edtech and entrepreneurship and lifelong learning. So I welcome all the panelists to come join us.

MAX: Good morning, everybody. I'm Max, I'll be moderating this panel. Very happy to have you all here. Awesome. So I think Nathanael left us on a fantastic note at the start of his talk, he said two things. First, he said we should try to anticipate the expected harms. And then he immediately said, don't just focus on the negatives. So I think this talk, building AI ethically or building ethical AI places us right in between that with ethical alignment.

So with that, as was mentioned, you guys all come from a rather eclectic backgrounds. And so I think an interesting place to start this conversation would be talking about how you all got interested in the ethics of AI. And, Richard, I'd love to start with you, given that you've kind of been all over the map in academia and in the private sector.

RICHARD: Is that on?

MAX: It's on.

RICHARD: Yeah, good. First, I want to thank James for this opportunity and privilege to be here. It's been a very informative two days. So thank you very much, James, as my colleague doing this thing called a MRPL degree at Harvard Divinity School, which is a one year career oriented degree.

It's a big question how I got interested in AI ethics. I guess there's two sources to that. The first, I went to an all boys Catholic high school back in the day. And while it was very good at teaching things like rhetoric and academic subjects, I never learned to type. And fast forward to my time at Nokia, where I was a senior vice president. They kind of innovated in texting, and I was terrible at texting, so I got in my head that there must be a better way to input data.

And I decided to co-found a firm that basically pioneered voice recognition technology. Long before Nathanael was involved in the venture capital business of AI, before we even talked about it as AI in and actually in a company here in Boston called voice signal, which was then five years later sold to a company called nuance in a bidding war with Microsoft and Google.

And it was a very lucrative deal for me. And so I went off and did some other things. But it got me very interested in AI because the voice recognition which used what you call hidden markov models was the basic approach or method that you used. And so then I went on into a lot of different firms in the I area, kind of simultaneous with that.

I saw an article in the New York Times Magazine at one point, which had the title, "Gravity is an Illusion." It doesn't exist. And it was by a physicist called Eric Wehrlein, a Dutch physicist who introduced me to the idea of emergence. And I got very, very interested in emergent phenomena and consciousness.

We talked about the mind, life, cetera. This idea that you can go from order to chaos in a spontaneous, automatic way. And then when I started thinking about AI, I put that together and said there's a real set of emergent phenomena that are occurring with AI where you get things that you don't expect, you don't anticipate, you don't plan. And so that had got me concerned. So it was both the practical thing, as well as the existential element that concern me about AI.

I now sit on a bunch of AI boards, a public board that's focused on image processing, Chairman of that, a robotics, AI company. And I am very concerned about the chaos that some of these companies are facing in terms of development, design, and employment and some of the things that are occurring. And I appreciate Nathaniel's position that you really want to encourage the startups. And it was refreshing to see the thoughtfulness that you were using.

That's not always the case, though. There are a lot of people out there that are actually looking for help. And the harms are all over the board from basic bias in the infrastructure in our health system, in our economic system, in our welfare systems to the manipulation of user behavior and purchasing behavior and stereotyping to the basic safety issues around things like smart bombs and autonomous vehicles all the way up to the very serious existential threats which I think are associated with things like the intelligence explosion, where AI not just becomes a general intelligence, potentially, but something that's more intelligent than we are, which in my mind is very likely to happen at some point, although I don't know when.

So all that's to say, I became very interested in AI ethics, decided to put my career on pause, come up to Harvard Divinity School, do a one year master's degree with a thesis on AI ethics. So what I'm trying to do is put together a general framework that captures all these harms. But to Nathaniel's great point, also focuses on what can be done specifically, right? How do we-- it's moving fast, and we're kind of at a precipice. Unless we get our hands around this, I'm not sure where we're going to go.

So we need to get our hands around this. And there are specific things that we can do. And so that's what I'm trying to bring about. And my role on boards and in consulting, I'll hopefully be able to contribute to that. Sorry for taking so long with that answer, but it was a big question.

MAX: No worries. Yes, it was a big question. Jenn, I'd love to move to you next. The reason being I think one of the interesting perspectives that you bring us is that not only like some of the other people on this panel, you've worked in the private sector and thought about market forces, which has been a theme today. But you are also working on the global scale now with the UN. And so if you could help potentially give us a broader frame as you explain your interest, that would be great.

JENN: Sure. Hi. I realized that my bio online or if it was written by AI, then clearly I'm not updating something online about myself. That's true, I currently work at the UN in the development program at their chief digital office. And I work on the AI hub for sustainable development and also trust and safety.

My storied career prior to coming to Harvard Divinity School, was that I was the head of platform integrity for Facebook. And so much of my career over, I would say, almost 17 years has been to look at the ways that systems we use around content moderation or the other things that basically thwart and try to protect people from being victimized online. And hit a bit of a crisis of conscience. I had been working at it for so long that ideally, I'd like to take this posture towards problem solving, that I should work myself out of a job.

If I'm doing my job well, then the problem should be getting better. I don't want to be continuously working on the same problem over and over and over again. But I found myself in my career saying I was getting promoted and had a bigger team than ever. I was hired partly because I understood groups and I understood ecosystems, and I could see how it is that harms were being created by the very platform vulnerabilities that were out there.

But the problems were getting worse. Like I had access to all this data that Meta has. And it pointed in many ways that I was failing at my job. So I wondered if the systems and things that I had implemented prior were the wrong ways to problem solve. And I think we see these problems still getting worse. And so I felt like I needed a break to pick a place that was so far outside of the Silicon Valley. And I found myself here, which is perhaps not far enough when you think about how we're talking about market forces and things like that.

I was like, are we having the same conversation? But Divinity School, I thought, oh, I could end up at Kennedy School. I mean, I still took courses in there and stuff, but I wanted to really fundamentally different way of thinking about things. And what I got out of this program here was so different than what I expected.

But I think some of the driving motivators were the same. I think what I didn't quite understand is that what I was looking at is both a spiritual crisis on my part, but also because I was grieving. I question it. And I question because I'm not Christian, but I question coming here. And also the work that I do now with the UN is a means of atonement.

Looking at the ways that I perhaps tried to solve the problems of the past was so short sighted that I could not fully contextualize what the implications would be now, that the ways that I built systems had led to a certain kind of power accumulation that we are now reckoning with presently, that even my minor critique of the presentation is that we haven't fully really addressed that in these ethics. Whose ethics are we talking about? And ultimately, in these systems of ethics, inherently every system of ethics.

And it's so interesting being able to study religion because there's nothing in our moral or ethical language in many ways that hasn't stemmed from a religious background. But there is always a privileging. There is always an inherent in ethics. You have to ask ourselves, who is being centered in these ethics, who is being privileged in them, and who essentially becomes collateral damage, or who is being diminished as a result?

And so I think that coming here has been like a good lesson for me to consider that, and also to look at the ways that I was problem solving. And I can say that at the time, compassionately, I think I was doing the best I could at the time, but it was a bit shortsighted. And that inherently, the systems that we still continue to replicate and recreate are replicating histories.

These are not entirely new human problems. That, for example, trust and safety, and I believe AI as well, are carry colonial constructs of problem solving. They are not Democratic systems. And we shouldn't pretend that they are. And I also like I advise like Senate hearings and other things to say. Like we should just consider that you're asking companies to try to uphold certain types of systems of justice without asking them, is that even what they were constructed to do?

And also, even when we ask ourselves questions of justice or fairness, are we talking about retributive justice? Are we talking about restorative justice? Like what kind-- like there are layers to these things. And I think that coming here has been a good way to examine that more closely. And, and so yeah, let's just say I've had a storied career, but yeah, it's been great to be here.

MAX: Awesome. Yes. I think one of the themes that we've heard as people who've been trying to keep up with AI is this idea of becoming a lifelong learner, and you've obviously come back to school to learn more. Along that theme, John, you work in the Graduate School of Education, and you've been working on how to fuse AI with education. So I'd love to learn a little bit more about what your origin story is in that regard.

JOHN: Sure. Thanks. Thank you. Well, my PhDs in philosophy. So I either know everything or nothing. But I focused on foundations of math, philosophy of science. And they said, if you're going to do philosophy of science, you need to know a science. And at the time computer science had the fewest number of courses. So I took every one of those.

About 10 years later, I was at MIT in what was then called the division for study and research in education, that ultimately, most of those people ended up in the Media Lab. I left to start a company because I was really tired of the academic, let's think about this and how can we really make change and do something? And so when the question of when did you first come across AI, on some levels, it was before it existed as a topic.

But in the mid 80s, I was at a think tank here in the Boston area, and we developed an expert system that could solve linear algebraic equations. And part of what we were trying to figure out is how do we put this in the hands of the student, not subject the student to what the expert system can do. And to me, that fundamental question of empowering the user or empowering the student is how I want to think about uses of AI today. And it's persisted through all of my career.

When I-- for a while, I was at Turner Broadcasting CNN, and I was responsible for all of the education outreach. And for us, the question wasn't, how do we take Turner materials and put them on top of a student, but rather we created CNN student Bureau, where we trained high school and college teachers around the world, so that their students could produce short 2.5 minute pieces that could appear on CNN.

So again, the whole issue is how do we empower the student and how do we give this to the student in a way that allows them to understand the process. Interestingly, I think I'm in a project with Jenn because I've just gotten a contract from education above all, which is a Qatari company, and they want us to develop curriculum materials that they can distribute in the Global South that makes students aware of uses of AI, aware of social media, and how to think about what I call digital media literacy, how do we get the students again in control and not victims to what's coming through on AI and in social media.

MAX: Excellent. OK. So then also lastly, James, we were talking right before this, and I was asking you about the relationship in your life between spirituality and all of this AI stuff. And the first question you asked me was what is spirituality? So perhaps I should take a step back and let you frame your own story here. But I do think it would be awesome if you could incorporate your background with Buddhism and spirituality, especially given that this is the Divinity School.

JAMES: Sure. Well, I first got involved in technology-- I mean, an old science fiction guy from the first book I learned to read when I was in third grade because I had ADD was Tom Swift Junior, which was like Elon Musk on steroids. He was building rockets to fight the Russians and things like that.

And then I've been an avid consumer ever since, but I got involved in technology ethics in graduate school and in particular bioethics because of my interest in human enhancement topics, genetic engineering, things like that. And at the time, and hooked up with Nick Bostrom and the transhumanists and was kind of participant observer in the whole scene. But at the time, in the '90s, the people who were really enthusiastic about artificial intelligence or scared were the singularitarians, and they were kind of our millennialist subsect.

Some of them were apocalyptic. Some of them were, believing that the robot God, if you could build the right robot God, it would protect us from the Bad Robot gods. So I pretty much ignored the singularitarians and didn't focus a lot in our work at IAT on AI ethics. But my mind began to change in the 2000s around the time that Colin Allen and Wendell Wallach wrote a book called Moral Machines, which I found very it's short and very influential.

And their basic argument was that most human beings with the way that they approach ethics is not I'm a consequentialist or I am a virtue ethicist. They approach situations, particular ethical dilemmas, and they have more or less ethical competence to understand the various features of an ethical situation, relate to some deontological questions, some virtue questions, some consequentialist questions, a history of debate around these things.

And their argument in the book was that if we wanted to train AIs to be moral machines, that the way to best do that to approach that is casuistic education, to understand all these different moral features and all these different kinds of ethical cognition. And I found that kind of developmental approach very appealing, the notion that you couldn't just program Asimov's three rules of robotics in and have it work flawlessly, which was Asimov's point in the first place.

And so this kind of developmental approach then I think is very tangible. Now that we have these AIs that are being red teamed by folks in don't ever tell anybody how to make a nuclear weapon or don't ever tell anyone how to make a bioweapon. Also, the red teaming around what kinds of-- how to interact with people. And we've seen now with grok that you can get an AI that will be as racist as you want it to be or as misogynist.

So we're doing some of this ethical training, and we're also seeing that some LLMs have intrinsic ethical orientations already because of the way that they were trained in the corpus of things that they got trained on. And that I don't think that is going to-- I don't think that the intrinsic left liberal bias that has mostly been identified in the alums is going to last very much longer.

I'll just give you the example of deep sea. As soon as the deep sea set model was open sourced here in the United States, people said, well, you don't want to talk about Tiananmen square. But then they cracked it. The jailbreak broke it so that it would talk about Tiananmen square. And in the same way, grok shows that you can go in a different direction with these algorithms.

So more recently, I've been thinking more about the ethics of digital agents, because I don't know, I don't have a great track record as a technology prognosticator, but it seems like we're on the verge of people being able to build, and many people wanting to build digital agents that are simulacrums or digital twins of themselves, to do various kinds of tasks. We may have one for work and one for family and one for games or something like that. Or we may just have one that can do all these different kinds of things.

And my interest in the spiritual connection, I think is well, how can we use these kinds of digital agents to improve our lives? So the question earlier, instead of me scrolling through TikTok all day, which may in fact be what drives the dopamine. If I have a sense of dissatisfaction about the fact that I have digital addictions, can I program my digital twin or my digital superego to reduce or to discourage these bad habits on my part? And would there be some kind of measurement function that you could say my eudaimonia should be maximized, as opposed to my simple pleasures, my hedonistic pleasures?

So I'm very interested in that kind of moral enhancement aspect. But my final thought would be that it's very difficult to be a moral person in an immoral society. And the society that we're in is one where oligarchic monopoly capital has captured the state and is quickly dismantling it and has changed our, at least for the moment, changed some of our social values. I think it's difficult for me to imagine that tools designed by the people who control this particular social order would necessarily enhance human beings in the way that I want them to be enhanced.

For instance, digital democracy. I think one of the things that's a problem with our democracy is how difficult it is for people to have an episteme that reflects the world accurately, and then to know how to affect it in a political way. And I think digital tools could-- digital twins could be one of the tools that we use in that space. But is Google or Facebook, X going to be the designer of a true digital citizen twin, that it's going to change society and empower the citizen as opposed to the oligarchs? I'm kind of skeptical at this point. So a number of concerns.

MAX: Gotcha. Awesome. OK, so now I'd like to get a little more specific in the conversation. So I'm going to ask two questions. The first is going to be addressed to John and Jenn. And the second, I'll come to you too, given our backgrounds here. So the first would be that-- two themes that have come up so far are human empowerment and bias. And in some ways, these are at odds with each other. The reason being that one, we don't have one universal ethical system. And so if you build that any system into a system at the outset and the bias carries it, it can have all sorts of different ripple or second order effects, which will be what the second question is.

So John, I'd love to come to you first and ask the question specifically in the context of education, how can we create educational systems that ensure that we're enhancing critical thinking and creativity, as opposed to creating or optimizing for standardized learning in our students?

JOHN: Well, I think the answer is you have to try. If you're starting out saying there's this body that you have to learn, body of knowledge versus how do you approach getting knowledge, how do you figure out what you believe. So, for example, part of what in this project we're thinking about is where do you get your news from? Go to your social media. Take a look at how that contrasts with what the BBC says. How does it contrast with Al Jazeera? How--

Take a look at multiple sources. Think about multiple perspectives. And to me, this is a path of inquiry. It's not getting to memorize facts. And I think that to me, when I think about the student constructing their reality, constructing their knowledge, it has to begin by looking-- getting them aware of the sources and trying to figure out how to vary what they're learning.

We're in such an echo chamber today and a series of echo chambers, none of which talk to the other. And I think if we can get students to recognize that they have fallen into this small area, then to me, that's the purpose of education. It's not to give you 17 more facts. It's to get you to start thinking critically. I mean, so we use the word critical thinking way too loosely. To me, that in many ways is the purpose of education.

I have a question, I guess, in the inherent question. When you're asking about bias, what is it predicated upon? And I think that my question around that is because it assumes to some degree, some level of centralization that everybody is going to be using the same thing.

MAX: Yes.

JENN: And so then really, it's an appeal to a certain group of people about how it is that they're choosing to problem solve and choosing to make something that you believe that everybody's going to be using. I think that something that's been interesting about being at the UN is really having to immerse myself in what does it mean to have the right of self-determination.

And that relationship to AI I'm not sure what that's going to be, but I'm saying that the challenges around AI is that it fundamentally in its power accumulation, is taking away our sense of agency. I asked this question yesterday because I'm not entirely sure what to do about it. I'm not saying I have an answer for it, but I am saying that it's worth the provocation. Because if we believe that there is an inherent human right to self-determination, how many people believe that there is a right to self-determination? How many people believe that they are relinquishing parts of that with their use of technology? A lot of you.

And so there's a tension that I want to just call out because we can keep talking about bias. And I appreciate what you said earlier because let's talk about the harms but also like because the bias is the effect of it, but actually get to the root cause of it that is predicated on certain assumptions about how we're trying to problem solve and how we're trying to maybe dismantle how we understand these relationships. And I worry about that as much as I'm investing in things like AI for water systems and health and all these other things that the UN is doing, because I believe that those are also fundamental human rights, this one around self-determination is puzzling for me.

And I sometimes think the question isn't it shouldn't necessarily be about bias, but like what is that relationship around bias? And does that actually take away people's ability to navigate the world and have self-determination? And so yeah, that's kind of where I land on this one. I'm not sure if there's another part of this question, but--

MAX: Sure. Yeah, just I guess to coax a little more out of that. On the self-determination front, is the idea that the I is giving you information that may or may not be angled in a certain direction. And so in that way, it has influence over you and takes away your self-determination. Is that the idea?

JENN: Yeah. I mean, so in my experience in building systems, and I think we also have to recognize that AI has pre-existed the age of LLMs, that so much of the so much of the world in my world, the trust and safety, for example, is that everyone else in the world is downstream of market factors because early adopters ultimately determine what is respectable.

And so-- and what is healthy and what is desirable. The challenge with things like AI, at least if we look at LLMs, for example, is it gives you an answer, but there's no appeals process. With the way that trust and safety used to work, it used to be that at least you could say, oh, why did you take down my content? I believe I have an inherent human right to communicate myself because I have a freedom of speech, or I believe that there's a right to freedom of speech. And so I'm going to appeal this.

And that helps us course correct a little bit. AI systematically doesn't actually have that inherent sensibility around its own rights. It'll just say like, oh, this is an undesirable answer. Therefore, I will just try to mitigate this and not serve this to you. Even if maybe that it would be a good answer for someone else in a different context and using the same kind of prompt.

And so I've just considered the downstream, the bias does have an impact if we believe that everybody will be using the same thing. But again, we have to contextualize what that means. And I also think that we have to think about, are the systems that we're building leading towards improving the human right to self-determination. Like how do we course correct for that? And yeah, and I think that we should be problem solving for that. But again, I think there's a lot of ethics that aren't inherently built into the ways we're trying to problem solve these things.

And so yeah, bias becomes a problem. But even the positive like thinking about the moral like cultural violence aspect of it, we can all say that it's good to screen certain answers. But again, good for whom, in what context? And we find this a lot with human health answers. What is considered to be healthy in Uganda? What is needed in that particular context is not necessarily the ways that we would optimize for health in the US.

We have different access and different solutions and different kinds of medicine and different kinds of cultural values. And so to have everybody have to use or the wide adoption of the same raw models that are just kind of in the supply chain of things adapted for a market, but inherently have some of the same issues, even when you talk about jailbreaking or red teaming, those are so far down the supply chain that everybody is still inheriting the evils that were before it. And so we're kind of not totally solving for it.

But yeah, maybe my answer-- more questions than answers.

MAX: Yeah, I think, John, did you have something to add to that?

JOHN: Well, to me, so much of this goes back to 1948 and Shannon and Weaver's mathematical theory of communication, which laid a foundation for so much of what we're doing and what evolved out of that was media literacy questions of there is a message coming, who sent it? What are they assuming about the audience? What are they assuming about you? What can you believe about the message? So it's not just that messages arrive, it's that they're sent.

And when it's coming from an AI, because you're having a conversation with it, that you have to be aware that this isn't some neutral truth that's just happening. Again, you have to think about why is it being sent? What is its assumptions about the audience, and how do you react to that? And it's that perspective that to me makes this much more critical in terms of someone interacting with an AI.

And I think there's a lot of talk now of connecting the new AI with the GOFAI, the good old fashioned AI. And I think those kinds of connections are talked about, but it's pretty confusing to me how that would actually come about.

MAX: Awesome. So I'd love to now move from talking about AI at the individual level to a broader level. So I'd love to go to you next, Richard, if that's OK. So you've obviously had extensive experience as a CEO. And as the CEO, you help manage a number of people who are individually working but in a large corporation. And so these effects, and you don't have to necessarily speak to bias, you can speak more broadly here, of AI and the impact that can have for empowerment or anything else at the individual level, how do you think we should be managing this as corporations?

RICHARD: Yeah, well, that's a great question. And I am very concerned about bias in-- of our artificial intelligence systems. It's been well documented. But you can see biased outcomes, which create a situation which marginalizes communities. And you can see this in our health care system. You can see it in our policing system, in our judicial decision making systems, in our educational system, all throughout the infrastructure in society, there's a bias, and it leads to harms.

And for companies that are involved in the use and deployment and development of this AI and the systems, there are definitely specific mitigating actions that you're able to implement that will help prevent that. And right now, really there's a lot to be done in that area. The AI is penetrating society at a much more rapid pace than we are putting in guidelines in place and putting in procedures and policies within corporate environments that assure that our AI is not biased.

There's things you can do, right? Very specific things. There's participatory and inclusive design methodologies that you can incorporate into your development team. There are third parties that you can work with, like data for Black Lives, because a lot of the bias, by the way, is people should understand is created by biased data, which is data collected from human beings who lived biased lives and had biased behavior. So the bias is built in.

What's really insidious about this bias and harms created here is that they're somewhat invisible. And this gets to Galtung's framework because there's these beliefs and ideologies in society that says-- and I'm astonished sometimes at how many people embrace this idea that AI is neutral, that it removes human prejudice. It creates efficiency. It's great for human progress. Absolutely. But it's not neutral. But that belief that it's neutral kind of hides the biased outcomes that it can create.

So I think there's either you can diversify your data. For example, within an organization, data diversity is a big deal. You can diversify your development team and you can incorporate, for instance, community input to make sure that your bias from your data is removed. Similarly, there's things you can do in the algorithmic development, where you create feedback mechanisms and checks.

You can embed values in various ways into the algorithms that prevent this. So the question is, people kind of know about this and companies kind of know about this. But AI is progressing so rapidly and getting embedded in our society so quickly, that when you weigh it against the profit motive and getting a product out the door, it you lose the potential for protecting yourself. And there's no rule that says, AI is going to lead to a great outcome for society. It's really up to us to implement these specific techniques within organizations to ensure that the outcome will have our progress, but will also create just society which is really in jeopardy.

I think that the potential impact of AI is greater than the printing press. Someone mentioned the printing press is greater than just about any innovation or transformation in our society, but we really need to harness that in a way that protects society because it can completely reshape society. And the danger is the collapse of human autonomy and human agency. Right? And that's why we need to really be diligent about how we go about the development and the deployment and use.

And we need to educate people. This is a little pet peeve of mine, too, because I'm at the Harvard Divinity School. But honestly, there's a little bit of an attitude that don't use AI, and you're sticking your head in the sand. And these are people who need to be responsible citizens that we're training. And they need to understand the pros and the cons and how to use it well, and the dangers of AI.

And the only way to do that is to create an environment where they're using it as a tool for good. Right? And to say, I'm afraid people are just going to be cheating, so don't use AI is not the way to approach this. I think actually, the Ed School has a much more innovative pedagogical approach where they're doing things like debates and group sessions where they encourage people to talk and converse, as opposed to just producing a lot of response papers to major readings.

So I think we here at the Divinity School need to rethink our approach to that, so we can create citizens that are responsible and understand the dangers, who then can turn prevent those dangers. So we can harness the power of AI for the good at the end of the day.

MAX: Fantastic. So James, I'd love to go to you off the back of that. And just to help angle it a little bit, one of the things that Richard was talking about is deploying AI. And the other aspect would be that we obviously want it to be deployed. So equitably, and you've been a Democratic transhumanist for a while, so I'd love if you could explain what that is, perhaps to people who don't know what it is. And then, yes, whatever follows.

JAMES: Well, transhumanism as a movement had many different factions and wings, and the survey research that I did showed that most of them, or plurality of them, were left of center. It's hard to imagine these days when the few right-wing billionaires have become the poster children. But Democratic transhumanism, as I argued for it, we call it techno-progressivism these days.

That idea is basically that we want the benefits of emerging technologies. We believe that emerging technologies will actually occur. Things like life extension, artificial intelligence, genetic engineering, whatever. Things that aren't often being debated in public policy. And we think that there's probably a way to have better outcomes than worse outcomes, and that we're probably headed to worse outcomes unless we work on the better outcomes. So that's the basic nub of it.

So we tend to be very critical of left Luddites who think it's not going to happen or think it's worthless to talk about trying to make it better. And also the techno-utopian libertarians who don't think that you need to do anything about making it better. It'll just be better. So that's the basic idea there when it comes to the particular question of algorithmic bias, I have often argued and it may be coming less true now, but I've often argued that it's easier to fix an algorithm, a racist algorithm, than it is to fix a racist cop.

And that the racism in the algorithm is because of the history of racist cops, and we want to change that behavior. One of the ways to do that is to replace them with an algorithm making the decisions, or to guide them with an algorithm, making decisions, and then try to change that algorithm, so that it's not as racist. Now, one of the things that may be happening is that these systems are becoming so opaque and difficult to monitor.

There's a paper out that just did an analysis of the valuations of different lives around the world in the different LLMs that they tested. And the LLMs were, in general, valuing lives in the developing world more heavily than lives in the developed world, which I thought was a really interesting outcome from a kind of UN point of view, putting a thumb on the scale for the consequentialist outcomes of the developing world.

Is that a bias? Yeah, it's a bias. I mean, do we want to fix all the biases? I'm not sure we do. One affects all the biases. I'm sure we want some of these biases. If you had an algorithm that just told you the most likely people to pick out a blind in the TSA, it would probably be a pretty biased result. And we have never really done that. We've had more randomness in TSA because it would be politically unacceptable. People would get upset if we had one that was just picking people that were high likely suspects. So there are many things to go into the consideration of whether something is a bias or not. I think it's an open question.

JENN: Can I ask a follow up question?

MAX: Sure.

JENN: What makes us think that our respective ethics aren't biased? And also inherently are these-- when we say what we're doing something for good, like, I think my question is like, do we-- like, how do we qualify that? But I guess it's like is our biases inherently bad? Or the context dependent? But I am wondering I just want to ground us in the fact that in-- what's amazing about Divinity school is looking at the ways that people believe that the insidious nature of believing that your benevolence will lead you to doing good, and then realizing the cultural violence in it that actually sometimes what we believed was intended to be good led to terrible, dire consequences.

And I feel that certainly in the work that I did at Facebook. I want to believe that the things that I did were good. I think in general we trended that way. But there were also decisions that we made that were terrible, even if we thought they were good at the time. And I think only history kind of only like looking back at that, do you get the kind of 2020 vision and not even quite 2020 vision on our past decisions?

But I'd like to think we were trying to work ethically. I think, same to what Nathanael said, I've worked with incredible humans. I don't think it's that we inherently were bad people, but we worked within systems that often led us to make bad decisions. And I think about this sometimes in Hannah Arendt's reflections on the banality of evil, that sometimes you're just too short sighted to see that we might be a cog in a wheel, an operator, an engineer that in her storyline leads to the Holocaust, a contributor to that.

But in our storyline, what is that? And I guess I'm just curious about-- we're talking about these kind of things around bias, but I think we're talking around like if we can root this in a case study, I'd be really curious. But yeah, I think that my ethics are biased. I'm trying to accept that and try to still navigate through it, because I don't think I'd get rid of my ethics, but I just want to name that I think ethics are also deeply biased in that they are not inherently 100% good in every context with every person.

JAMES: Well, I mean, when ProPublica analyzed the criminal justice system of Florida's use of sentencing algorithms, they found that not only was there the intrinsic bias of the fact that African-Americans were rearrested at a higher rate historically and therefore algorithmically would be predicted to be worse worst candidates for parole, which is problematic. But also it was extrapolating from that and adding additional cautions to releasing African-American inmates over and above its algorithmic prediction.

So, then if you correct it, you have to say, well, that history, we're going to say, we understand why that history occurred, and we're going to have a bias of an affirmative bias to say we don't want to have a racial bias in sentencing algorithms. And I think that that's the role of ethics, as you say, having that kind of ethical bias in the world is the role of ethics.

MAX: Yeah, I think just so we can cut in-- so that we can get some questions from the audience. But I will say, I think this question of should there even be bias, because it can be helpful potentially, but it can also be super destructive. I think that's one that we can all think about.

SPEAKER 2: Sponsor, Religion and Public Life at Harvard Divinity School.

SPEAKER 1: Copyright 2025. President and Fellows of Harvard College.



 

 

 



 

 See also:- [ RPL Past Events ](/news/rpl-past-events)