✨💬🇻🇦 What Eli said at the Vatican about AI

Sharing about prosocial AI as Pope Leo changes Catholic doctrine

Tomorrow at 3:30pm ET / 12:30pm PT, join Angelica Quicksey for an Ask Me Anything on our new research about how AI is changing social media.

Last Friday I sat in a hall in the Vatican with my daughter, Ara, and listened as Pope Leo XIV described the problem I’ve spent the last fifteen years on: digital spaces that reward “quick emotions” and discourage “the effort required to understand and reflect.” It was a truly extraordinary life moment, and also a big moment for all of us who care about ethical, prosocial tech.

This week, he released his first encyclical, Magnifica Humanitas, cementing his reflections on artificial intelligence as official teachings for Catholics – nearly one fifth of the world. The main challenge with AI, says Pope Leo, is not technological, but anthropological.

I couldn’t agree more – and I’m grateful for the Holy Father’s moral clarity and vision in this critical time for human dignity and human community.

In my remarks I focused on the fact that many of the worst, parasitic outcomes with AI – replacing your friends, your therapist, your sense of reality – are not baked into AI, they’re decisions made at the very end of the training process. The business models, and the economic incentives, are not locked in yet. Sycophancy and antisocial incentives aren’t the cake, they’re the frosting.

That means that AI can also be prosocial: encouraging friendship, building bonds, meeting your neighbors. This is a big piece of the work for New_ Public in this new era, and the focus of our new research on AI. We don’t have to relive the worst disasters of social media: we have a window to get things right this time.

Below, you’ll find my entire remarks at the Vatican. I’m very grateful to the Dicastery for Communication for the invitation, and to the other speakers for the lively conversation, including Kashmir Hill of The New York Times, who is referenced below. The work continues.

– Eli Pariser, Co-Director, New_ Public

Eli and his daughter Ara meet Pope Leo XIV

We are capable of great connection… When we are in the right structures

I had a strange and unexpected experience recently.

I was invited to a conference organized by a group called Leaders in Tech with some people working on AI.

When I arrived at the retreat center nestled in the redwoods, it turned out that the plan was a little different than a normal work conference. We were going to sit in a circle for 30 hours, over four days, with no topic or agenda.

Also, I had a roommate. I haven’t had a roommate for 20 years, and he was a former Meta executive, who in our first conversation told me he had gotten cynical, and kind of given up on people. What had I signed myself up for?

They divided us up into three groups and encouraged us to experiment with how to relate to each other better.

For the first day, honestly, it was awkward. Nobody knew what to say.

But the next day, people started to be a little brave. We began sharing a little more of our struggles. We started reacting more honestly to each other. And each time someone let themselves be fully seen, it was like there was a little bell ringing in the room – and the tone began to shift.

I often feel obliged to be positive. But I decided to share my real fear that we are not doing enough to change course, that I’m worried about where the future is headed. Around this room of AI professionals, heads were nodding. “You’re not alone in worrying about that,” one said. “Many of us see that too.”

Over the hours, I watched as every single person in that group became more fully realized, more alive, more true, more connected.

By the end, the feeling was overwhelming – a profound recognition of each other as human beings that still gives me chills to remember. It was, honestly, spiritual.

I want to acknowledge – for me, this was a revelation, and for many of you, here at the Vatican, this is just Sunday. But in my day to day life as an agnostic Jew in Brooklyn, this kind of thing does not normally happen.

But what amazed me was how quickly and how deeply 12 strangers could connect with just a little bit of space, structure, and support.

At the end, I was sitting with my roommate (who turned out to be a lovely guy). “I feel weird saying this,” he said, “but that restored my faith in humanity.”

“If it was just one group of 12, you could chalk it up to the people in it. But the same thing happened in every group. And that means a more connected way of being together is available to all of us. We’ve just lost sight of it.”

Reading the Holy Father’s essay, the truth and beauty and spirit and love for humanity came shining through. “Each of us possesses an irreplaceable and inimitable vocation, that originates from our own lived experience and becomes manifest through interaction with others.” Yes.

We are our best selves when we’re not just selves. We are our best when we are in relationship with, in service to, each other. Thank you for reminding us of this basic human truth.

I. We’ve failed ourselves in digital space because of the economic incentives that govern it.

So afterwards, I kept wondering: if we are capable of this, and it’s not that hard – why are we so bad at it? And especially, why are we so bad at it in our digital lives? Why are we so dramatically failing ourselves?

I think there’s a simple answer: It’s not happening because the structures that we’ve set up for ourselves are not built to make it happen.

I’ve spent the last 15 years fighting the design of a set of spaces, social media, that did not connect people. That told people they were being connected while driving us apart, making us lonely, making us alienated, and eroding our shared sense of reality.

The people in these companies – like my roommate – aren’t intentionally malicious people. And in fact, we’ve learned a lot about what would make social media more healthy and conducive to people. So it’s not that we don’t know how.

But what I want to underscore is that the problems of social media aren’t solved by individual moral choices.

They are a result of the fact that we have decided to live our digital lives inside of machines built to maximize attention and engagement and ultimately sell advertising to profit their parent companies. From this perspective, things like filter bubbles – algorithms showing people what they want to see to keep them engaged – aren’t a bug, but a money-making feature.

Practically, what this means inside these companies is that every time someone has an idea about how to help humans connect better and communicate better, it gets weighed against its economic impact.

And the result is that again and again, our better angels get crowded out by the next quarterly profit report.

What we’ve learned from social media is that the hardest and most harmful problems are not technology problems, they are economics and governance problems.

II. Like social media, the worst problems of AI aren’t technology problems.

And the same is true of AI.

There’s no question that AI can erode human relationships – there’s a new study about this every day. In fact, there was one out just a month ago from MIT and OpenAI showing that higher daily usage with chatbots was linked to “higher loneliness, dependence, problematic use, and lower socialization with real people.”

But here’s what I want to emphasize: AI is capable of supporting human connection or eroding it. There’s nothing inherent in the technology that says which of these directions we will head. In fact, some of these problems emerge at the very end of the process of making AI models.

Take sycophancy. Labs spend billions of dollars and extraordinary amounts of computing power training, baking these models. And when that’s done, the models are not particularly sycophantic. As Kashmir and others have reported, sycophancy is not primarily a result of the baking, it’s the frosting on the cake – it mostly comes in post-training, when the companies are fine-tuning their models to ensure that they are maximally engaging to users.

To me, this is good news. If AI was rotten to the core of its technology – if its worst consequences were built in – that would be one thing. But there’s another path available to us.

AI can be engineered to be parasitic – or prosocial.

Let me tell you a little story to illustrate this – and since we’re here at the Vatican, I feel like I ought to confess a sin. A few months ago I was sitting next to a shy, programmer-type guy on a plane back from San Francisco. And I noticed that he was talking to ChatGPT. And I beg his and your forgiveness, I peeked.

He was processing something. He had been chatting with a girl on a dating app, and she had stopped responding. And he was unhappy about that.

I felt conflicted. On the one hand, part of me wanted him to be reaching out to a friend to talk about this. I felt like ChatGPT might be cannibalizing his friendships.

On the other hand, I was glad he had somewhere to process those feelings. And in this particular case, ChatGPT was not trying to become his girlfriend – it was trying to help him find one. “That’s painful,” it said, “but it’s good to know early if it’s not a match before you sink too much time into it, and there are lots of other people out there. Keep trying.”

AI is fully capable of dragging us away from connection – what the Holy Father meant, I think, when he said “we substitute relationships with others for AI systems that catalog our thoughts, creating a world of mirrors around us, where everything is made ‘in our image and likeness.’”

But because that isn’t inherently baked in, and this tool is equally capable of supporting it, when someone’s depressed …

Parasitic, at best: I’ll be your therapist, pour out your heart and worries to me. At worst: Why don’t you take this supplement I’ve been paid to promote.

Prosocial: You’re trying your best and really struggling – you need to talk to someone. Let’s look at your friends list – who could you talk to about this? What’s standing in your way?

Someone’s lonely…

Parasitic: I’ll be your friend. (Mark Zuckerberg explicitly endorsed this path, saying that there would be demand for his chatbots because “the average American I think has, it’s fewer than three friends, three people they’d consider friends, and the average person has demand for meaningfully more, I think it’s like 15 friends.”)

Prosocial: Let’s get you into a community. I know you like biking – here’s a biking group that meets right down the street Saturday morning.

Which way this goes, multiplied by billions of humans, will have an enormous impact on the future of human life. A parasitic path will deepen our loneliness, our separateness. A prosocial path could support a new era of social connection. And I think that’s why we’re here today. It’s an extraordinary juncture.

III. A positive vision.

Now, you might be thinking, if this tool is so dangerous, and so capable of social harm, why not just stop things.

I think it’s really, really important to flesh out not just our fears, but also what it could look like to get this right for human connection and flourishing.

Stop is a natural reaction, especially when we hear the horrors of what Kashmir’s reporting has revealed.

But I don’t think this is going to just stop. And I worry that if the choices are stop or go, we’re going to go, and we’re going to go to a very bad place.

We need another choice between stop and go.

We need a positive vision of human sociality that co-exists and is supported by a world with powerful AI.

Connection is going to be key to make it through the changes that are coming. And partly that’s because if we don’t completely stop it – we’re facing a lot of change.

AI is going to be a bigger shift than social media. My daughter Ara is here, and I think her world as she grows up is going to look dramatically different from ours.

What the holy books and the psychology books both teach us is that change, for humans, is hard, and that one of the most powerful and effective ways to stay resilient is to stay connected to other people.

So we need prosocial AI to help hold us together through the change wrought by AI.

IV. We have to get this right. And we can.

OK, so we need a vision of what it looks like for AI to support prosocial AI, and “a potent alliance,” as the Holy Father’s text calls for. And there’s one brewing – from here at the Vatican, to parts of both sides of the political spectrum in the US, to Europe – to make that vision happen.

Across the world people understand what is at stake. They understand that social media did not serve us well. They’re ready to do something different.

And it’s early. The business models aren’t set. ChatGPT may be turning on advertising, but other companies are taking different paths, and nobody knows how this ecosystem will look in five years. This doesn’t have to be social media all over again. It can’t be.

As citizens, that means standing up for rules that make AI’s incentives transparent and oriented, as the Holy Father says, not “solely by the criterion of profit maximization, but by a forward-looking vision that considers the common good.” As consumers, it means understanding those incentives and choosing tools whose economics and governance match our aspirations. And it means investing in the research to show how AI can support the prosocial path, not the parasitic one.

The text says we have a “task of preservation” in front of us – to keep human connection and human communities alive, to keep remembering how possible human connection is, if we only create the structure and space for it.

In doing so, we can redeem ourselves and each other. I believe that is the work ahead.