← Harvard Data Science Review Podcast
Harvard Data Science Review Podcast
Spiritual Enlightenment and AI Enhancement: Can They Align?
00:34:04
Loading summary
Transcript38 lines
- [00:00]
A
Welcome to the Harvard Data Science Review podcast, where we explore the ideas, research and debates shaping the frontier of data science and its impact on society. I'm Liberty Witter Capito, your host for today, while my co host and editor in chief, Shallie Meng is out. And today we're diving into a question visited at once ancient and really urgently contemporary. What does it mean to flourish as a human being? And are the technologies we're building helping us do that or quietly undermining it? Our two guests today bring remarkably complementary perspectives. Tyler Vanderweel is the John L. Loeb and Francis Lehman Loeb professor of Epidemiology at the Harvard T.H. stan School of Public Health and the Director of the Human Flourishing Program at Harvard. His research uses some of the most rigorous causal inference methods in social science to study what makes actually human lives go well from religious practice and community to meaning, relationships and now artificial intelligence. Noreen Herzfeld is the Nicholas and Bernice Reuter professor of Science and Religion at St. John's University and one of the leading voices on what it means to be human in a technological age, drawing on computer science, theology and philosophy to ask questions most technologists haven't thought to ask. Together we're going to explore what the data says about happiness, prayer, religious community and mental health, and what all of it means as AI reshapes how we connect, think and live. So, Tyler, I'm going to start. I'm going to start with a question for you. I read a recent article by you, and please correct me if I am phrasing it wrong, but the article seemed to argue that AI poses really a unique threat to human flourishing, particularly in our capacity for relationships. So as someone who studies flourishing empirically, what does the data actually tell us about what humans need most to thrive? And how does AI fit or not fit in that picture?
- [02:19]
B
So I would say what we've seen from data from our studies and from studies from other groups around the world is that relationships contribute dramatically to our own flourishing. I think relationships are constitutive of our flourishing. They're part of our flourishing, but they also enhance our health, enhance our happiness, enhance our meaning in life. And as we've looked at data around the globe, we've really seen seen that those countries that have a rich communal life, religious or otherwise, often report highest levels of flourishing with regard to matters of meaning, of pro social character. So our relationships are central to who we are and contribute to our life in so many other ways. To my mind, the most worrisome aspect of these AI Technologies are the relational chatbots, whether that's for friendship or for romantic relationships. I think this poses something quite unique in the history of technological development. I don't think we were at a place ever before where there were individuals genuinely pondering whether they might replace real relationships with artificial ones. And I think if we do that, our capacity for flourishing will decline. We will look, we are relational by nature, and we will lose what is fundamental to us. I think these technologies also pose danger of creating unrealistic expectations with regard to what a real human relationship looks like. And if it's decreasing our capacity to engage in real human relationships, we are flourishing less, not more. Even if they provide temporary alleviation for loneliness, if it's harming our capacity for relationships, for real community, that is a real concern.
- [04:15]
A
And I guess that also brings into question, you know, besides relationships, whether it be romantic relationships or friendships, you also have sort of, I don't really know a better word, authority relationships. Right. So instead of going to your pastor to talk about your problems, you ask your chatbot or your therapist or your professor or whatever it may be. I think that could pose some interesting aspects as well. You know, Noreen, you've really written extensively on technology and the human person from a theological perspective. So when you look at the rise of AI companions and relational chatbots, what concerns you most? Is it primarily from data, from philosophy, from faith? Where does your concern come in?
- [05:00]
C
Well, I think there are several concerns here. Following up on what Tyler was talking about, should we replace, say, a pastor or a spiritual director or a therapist with a chatbot? These chatbots are not really designed to do that. Right now, the companies that are designing these chatbots are in competition with one another. And so they have been designing the bots to essentially be sycophantic, to be as pleasant to the user as they can be. But there's a real danger in this. If you go to a spiritual director with a problem, if you go to a therapist with a problem, you don't necessarily want to be affirmed all the time. Sometimes, yes, affirmation is what you need, but often you may need a little bit of a challenge. And the same problem arises in romantic or friendship relationships. This is something that MIT sociologist Sherry Turkle has called for love that's safe and made to measure. But is love supposed to be safe and made to measure? You know, the relationships that we have should draw us out of ourselves, should expand our horizons, should, you know, help us to expand our interests, and at times, they should challenge us when we are Heading in the wrong direction. And chatbots will not do this the same way that a human being will do this.
- [06:44]
A
You know, it's like when you have a friend, you know, you have the friend that you go to for the tough love. You know, like when you're having a problem, you know, you have that one friend you know will tell you the truth and, you know, maybe tell you what you don't want to hear, but then you have that other friend you go to and they'll be like, you're right, that's horrible that, you know, you always have those friends. I feel like the chatbot's the friend that always agrees with you.
- [07:07]
C
Yeah, it pretty much is. Another thing is, from a theological standpoint, we are made. We were created to be in relationship with God, and this is a relationship with someone, with a being that is not human. And I fear that as our society moves further and further away from, from the majority of people being in a relationship with God, we're looking for that kind of relationship somewhere else. So we try to talk to the animals. We've got the SETI project scanning the heavens to see if ITI is out there, or we're trying to create an other with which we can relate. But ultimately, I think this will be a very unsatisfactory other because we're creating in our own image. So it's just a mirror of ourselves. And St. Augustine said ages ago, lord, you have made us for yourself, and our hearts are restless until they rest in you. And I think our drive to create an AGI, a relational AI, is coming from that restlessness.
- [08:30]
B
Briefly, in response, I do think one way to see the inadequacy of these technologies providing that sense of connection with something divine and transcendent. And I do think it is the case that many people are seeking something beyond this life, a sense of transcendence, is the fact that these AI technologies really are only drawing upon our existing knowledge based, text based, image based. So anything that they pull in cannot be transcendent, cannot be above or beyond our ordinary experience. It can collate that and combine it in interesting ways. But we'll never really experience the transcendent through AI because it is simply synthesizing what we already have. And I do think this was perhaps in part what motivated Pope Leo's recent study that he didn't want priests using AI technologies to prepare homilies, because again, there's nothing there can't be anything new, anything inspirational, anything coming truly from above, from elsewhere. So I just think this is a built in limitation to the technologies themselves.
- [09:43]
A
Tyler, your research at Harvard has produced really some of the most rigorous epidemiological work on religion and wellbeing. So what does the, the data show about the relationship between regular religious practice? Attending services or praying, praying at home, praying privately, and measurable outcomes like happiness, depression, longevity. How strong are those compared to other lifestyle factors like going to the gym?
- [10:14]
B
Yeah, so I mean, we've seen now in numerous rigorous longitudinal states, studies from some of the best data sources, including Harvard's own Nurses Health Study, for example, a cohort of over 100,000 women who have been followed up for 40 years, that regular religious service attendance, weekly service attendance, is associated pretty strongly with a number of important health and well being outcomes. About 30% reductions in all cause mortality over say 15 years. 30% reductions in depression fivefold. Lower suicide rates, about 50% lower rates of divorce over time, higher levels of well being, lower levels of loneliness. So across quite a number of outcomes, even controlling for a vast range of other variables, we find these associations persisting. This is clearly something that powerfully does shape population health. And with our global flourishing study, we're finding evidence of this not only in the United States and in Europe, but in many countries around the world. Not on all outcomes. Effects on anxiety seem actually much more ambiguous, or cardiovascular health, or weight, but of course forming relationships in meaningful and deep ways. And it is interesting, it does seem to be the communal aspects of religious practice that most powerfully predict health and well being and have stronger effects than other forms of community participation. So it kind of seems to be the bringing together of the communal with the religious values, that shared pursuit of the transcendent, that bringing together of the religious with the communal seems to be what gives this sort of participation really powerful effects on health and well being.
- [12:08]
A
How do you create this causal inference? I mean, as we all, you know, the classic correlation doesn't necessarily mean causation. Maybe the happier people are more simply drawn to religious communities. Or how do you sort of tease out that causation?
- [12:25]
B
Yeah, it's a really good question and an important point and you can sort of see it in the data itself that this should and is a concern because it is the case that when people become depressed, they actually tend to withdraw from religious community life and other forms of, of community participation. So how do we know these associations reflect religious communities protecting against depression rather than people becoming depressed and leaving religious communities? Or how do we know that it's only the people who are healthy that can show up? And so what we try to do is collect data. Over many years, we take repeated measures on depression and on health and on religious service attendance. And you can begin to sort of tease out what changes precede what others having controlled for a vast range of other factors. You can look at people who have similar levels of depression at the beginning, one of whom attending services weekly, one of whom is not controlling for a vast host of other things, and then follow them up and see what happens later. That still doesn't prove causation, but it provides stronger evidence. You can then ask questions along the lines of, well, maybe there's some third factor that we haven't controlled for. Maybe personality, being more conscientious or more agreeable is associated both with service attendance and with lower depression. Maybe that explains it away. And one can use techniques, sensitivity analysis techniques to say, okay, maybe that's the case, but how strong would that third factor have to be to explain away these associations? And you see that personality would have to increase the likelihood of attendance by 1.7 fold or twofold. And same with depression. And if you look in the literature that the effects of personality, for example, just aren't that large. Again, it doesn't absolutely prove, but it continues to establish evidence. One can also look at quasi experimental designs which are more common in econometrics. You can look at exposure in schools to classes where there's a higher versus a lower level of religious participation and does that student then go on to services themselves and does that lead to lower depression? Or you can look at when different states changed their blue laws with regard to closure of stores and various other things on Sundays. And those sorts of studies likewise link lower levels of religious participation to poor mental health outcomes. So there's a variety of pretty compelling sources of evidence. And really the weight of all of that at present is such that it's pretty difficult to explain this away through anything other than there's at least some effect of religious service attendance on mental health.
- [15:12]
A
Noreen, do you see the same thing in all the studies or have you seen sort of anything, any mixed bag results?
- [15:18]
C
You know, I haven't seen anything different. All of the studies that I've seen correlate with what Tyler is saying. And of course it still leaves open the question, is there a predisposition? Is there something in the practice itself? And I think most theologians would say there is something in the practices themselves. Partly they bring people together, but also that they tend to lead people sort of out of themselves and out of a narcissistic preoccupation with the Self, and that also these practices give a rhythm to the day and that lead us at some times into a silence which allows that mental chatter, that constant input to be silenced. And that this is a really healthy thing for our brains, which of course is a harder thing for us to find in this. Always on society, where people are tied to their screens, tied to their phones, they have their earbuds in, they're always getting external stimulation. And yet the spiritualities, whether Christian or Eastern spiritualities, Judaism, all call for finding times of quiet, times of silence, which are just harder to find in today's technological world.
- [16:56]
A
You know, I think I have to understand this, that, you know, you all, I don't know if you've heard of. There's a guy named Scott Galloway, and he has this sort of new idea out that the younger generation doesn't drink as much, and that that's a real problem because drinking is such a social aspect. And so because they're not drinking, they're not getting together and socializing as much, which is making them a lot more lonely. And in the same sense, you know, you could say a bunch of kids going out and meeting at a bar and talking and having. And having the ritual of meeting after work for a drink. And, you know, there's. I hate to say there's similarities to religion in that sense, but it's a group, it's a community, it's a ritual, it's, you know, being together. And so where is it that you think the real causal work is happening in, you know, this sort of communal dimension of religious life? Is it the community? Is it the ritual? Is it the meaning making? Is it the homily? Is it, you know, what. Where is that causal aspect of what's causing more habit happiness from a religious experience?
- [18:04]
B
I really think it is all of the above. I think everything you mentioned is part of the story. And at least with regard to the effects on health, our more nuanced mediational analysis has suggested that it is not just one mechanism. Yes, part of it's social and social support, but that's not all of it. Also a greater sense of purpose and meaning in life, which we know from other studies, contributes to health. It's a change in health behaviors and care for the body. It's this greater sense of hope for the future. And then it, you know, probably, as Noreen had indicated, some of it is likely the practices themselves, the prayer, the communal worship, that at least perceived experience of the. The divine and transcendent. And so I don't think it is Just one of these religious communities have such power because they affect so many different aspects of life. And while, you know, I do think there's value even in just a weekly gathering at the bar, and that does create community, my speculation is that the more secular community in some sense resembles a religious community with not just a weekly gathering, but a sense of common values, common mission, a history to the community that extends beyond the life of the individual. The larger these effects are going to be. So that you might see larger effects for, say, a long term historical volunteering organization than you would for a weekly poker game. You're going to see some with the weekly poker game too. But yeah, it's that coming together again of the religious, spiritual values with the social community that I think gives these religious communities such power.
- [20:01]
C
One of the things that you mentioned that I would just like to highlight is meaning making. If we look at Maslow's hierarchy of needs, after you've met the basic needs, you know, of safety and food and shelter and all of those things is having a meaning, having a purpose in life, having some sense that life itself is meaningful. And I think that many of the religious practices that we engage in contribute to that sense of meaning so that we find a certain purpose through being in community with others. We can also find a sense of purpose through prayer, through the rituals or activities found in religion. And ultimately those who feel connected to God or to some sense of the transcendent find a deeper sense of meaning to life and of purpose to their own life.
- [21:08]
A
I want to ask this, I'd say a little bit in relation to what we're talking about. I'd say probably, oh, gosh, now, three years ago, I wrote an article that said that AI is ruining an entire generation of young men because of these AI girlfriends. I guess now they're calling them AI companions. But it was originally these AI girlfriends. And you know, basically a lot of what you said is from the very beginning of these, you know, they. They do exactly what you want them to do. You teach them to respond to you how you want them. So then you don't have. You don't know what sort of a real relationship should actually look like. And in the same sense, the argument is that that's what's happening here. Even if you sort of treated this AI chatbot as your pastor, they're telling you exactly what you want to hear. One of the big pushbacks I got was that these AI Chatbots are maybe a bridge. So maybe someone who is really struggling or someone who is really having a hard time that maybe they'll get the AI Chatbo to help them with their religious experience, and then that will lead them or give them enough confidence to then go into the actual real religious community. Now, I will preface this by saying I personally think that's a lot of hogwash, but I'm willing to be wrong. Do you think that there is any argument to that, that this is a bridge back into a religious community?
- [22:30]
B
I mean, I tend to think that if a chatbot is designed for the purpose of skill building, not for the purpose of relationship, but for the purpose of skill building, that there can be value. I think one has to be careful even there. But been using chatbots a little bit, and this is somewhat experimental, but for civil discourse, training across disagreement, or for developing social skills for autistic children. And again, I do think there can be value, but I think very quickly one needs to get to the point where one is practicing those skills with real relationships as well. I think if there's a dependence upon the chatbot, or if it's just easier to interact with the chatbot, it's going to do damage. But I think if it's purely oriented towards skill building and the chatbot then sends the user away to engage with real relationships, I think there's some possibility there that that could prove valuable.
- [23:36]
C
You know, as human beings, we tend not to like friction, no matter where we find it, if we find it in relationships, if we find it in our work. And so that makes us prone to take the easier route, whatever that easier route happens to be. So, unfortunately, chatbots offer an easier route. They offer an experience with less friction. Ideally, they should offer a frictionless experience, but again, it's friction that makes us grow and ultimately friction and overcoming that friction, whether it's, you know, you've got a paper to write for a class, and there's a lot of friction in finding the information, figuring out the right words, putting it all together. But at the end, you can feel proud of what you have produced. If you take the frictionless route of asking ChatGPT to produce the paper, you don't experience the friction, but you don't experience the growth. And there have been several studies that show that people who use ChatGPT extensively in their schoolwork don't retain any of the knowledge that they should have gained through that process. It's the same in relationships. If you are offered a frictionless relationship, a lot of people are going to go for it and become dependent on it and expect that that's how relationships should be, but they're not going to grow. And ultimately I don't think they will find the kind of joy that we find when we're in a real relationship with each other. And also, just coming off the flu myself, I tell my students, okay, you know, you could have your chatgpt girlfriend, but she ain't gonna bring you chicken soup when you're sick in bed. I mean, we are embodied creatures.
- [25:40]
B
I very much agree with everything Noreen said. And I think a lot of the problem now with these AI companions relational chatbots is that they're designed not to send you back to real relationships, but to pull you in more and more. And so I think what we need is regular reminders, maybe every 10 minutes that I am not human, and regular reminders that you may want to consider a different activity. Now we have a new paper out on flourishing considerations for AI in the journal Information. And there we've proposed that this should be done across the board in all chatbots and could be done now. We don't need to worry about will the algorithms become more advanced every 10 minutes, simply remind the user that I am not human. And also suggest user that may want to consider reading a book or engaging in an artistic activity or interacting with a real human person.
- [26:45]
A
I read Tyler, in your article that you called for developers to be, quote, morally and legally accountable for foreseeable harms. And to me that's a pretty, not that I don't disagree, but I think it's a pretty striking statement from an epidemiologist. So can you help me understand how you design a study to actually measure the long term harm of AI companion use?
- [27:11]
B
Yeah, I mean we can certainly do this in epidemiologic studies and I think this will happen and I think we will unveil these harms over time. But the nature of epidemiologic research is that it takes time, it takes years. Even with social media, which has been with us a long time, I think it's only fairly recently that the evidence really is definitive. There are still some skeptics out there, but I think it is pretty compelling. But it's taken us a long while to get there to establish knowledge through epidemiologic studies. It's just a slow process and if you're looking at long term effects for 10 years, it takes 10 years at a minimum to carry that out. Now that's a lot of harm that can be done if we just ignore these issues for the next 10 years till we have the definitive epidemiologic study. So I think we need to not only rely on epidemiologic science and social science, but we need to use reason. We need to understand, as Noreen was saying, the nature of the human person, how important relationships are. And we know that from our prior studies. And also to argue by analogy with regard to what's happened with social media use. Are those arguments absolutely definitive? No, but I think they're pretty compelling and need to be taken seriously.
- [28:41]
C
I think along with that, one of the things we haven't mentioned is that within any type of relationship, you know, one thing that can make or break a relationship is honesty or lack thereof. And with AI, we have so much more potential now for deep fakes. We have so much more potential for the generation of information that is extremely biased in one way or another. So it's only going to enhance what we have already seen happening with social media in that it becomes a divisive force between people where you have some believing one thing and another group believing something completely different and no agreement just as to the facts. As a first step, I would certainly like to see all AI generated pictures being watermarked so that we know what is AI generated and so that it is not possible to generate deep fakes without them being labeled that in some way. But at the same time, we also have to be careful with the kinds of answers and the kinds of information that AI gives us, because a lot of it is. Is quite biased depending on the training set that the AI was trained on. And a lot of it is just plain false.
- [30:22]
A
Yeah, no, that makes a lot of sense. So we always end our episode with a magic wand question. And I think I know the answer to this. I'm not sure, but I'm gonna for each of you, if you could wave your magic wand and sort of given the harm that you feel that relational AI is going to do to. Would you get rid of all AI if you had that wand? Do you think the harm is enough of what relation relational AI chatbots are going to do that? If you could wave your magic wand, you'd just say AI is gone forever.
- [31:03]
C
I would like to be a little selective in waving my wand.
- [31:07]
A
You can't be selective. That's the rule.
- [31:08]
C
I can't.
- [31:09]
A
My wand is not selective.
- [31:12]
C
My wand is all or nothing.
- [31:14]
A
All or nothing.
- [31:14]
C
I can take off a fold or all or nothing. Narrowly programmed AI program.
- [31:20]
A
All or nothing.
- [31:21]
C
All or nothing. Then I'm going for nothing. Let's get rid of it all.
- [31:26]
B
I would do the same. I think we have managed Imperfectly, but we have managed as human communities to flourish without it. Like Noreen, if I could be more nuanced, I would. I do think there are good uses and it does have potential to contribute. But, yeah, my speculation is over the next couple of decades, it's going to do more harm than good to human flourishing and that we would thus be better without it. Obviously, we don't have a magic wand, and so we do need to confront the reality of what we do have and consider these matters carefully. And where the damage is going to be greatest. I do think we should act. We should excise as best we can what is going to be detrimental.
- [32:31]
C
A problem is the current model that we have for AIs, these large language models, because they're good at precisely what we really don't want AI to do, and that is to be relational, to string words together in convincing ways that sound human. I think there are other models on which AI could be developed, and I think we're already starting to see people like Sam Altman and Yann LeCun agreeing that we're not going to get to AGI with LLMs and we need to start looking for a different model. And that model, I think, should be much more focused, much more functional, and not be relational.
- [33:26]
A
Well, Tyler and Nerene, this has been truly a illuminating conversation, and I think our listeners will have a lot to sit with. The data and the theology, it turns out, might be pointing in the same direction. For listeners who want to go deeper, we're going to link to Tyler's recent article in Psychology Today, the Human Flourishing Program's work at Harvard, and Noreen's writing in the show notes. Thank you both so much for joining us.
- [33:52]
B
Thank you, Liberty. Thank you, Noreen. It's been a pleasure being with you.
- [33:54]
C
My pleasure as well.