WEBVTT 00:00:04.080 --> 00:00:12.420 Welcome everyone to this evening's installment of Divine Mercy University's webinar series for mental health awareness month. 00:00:12.720 --> 00:00:17.815 I see, people are still gathering, joining the call. 00:00:18.475 --> 00:00:23.935 I'm going to, share just a little bit about our presenter this evening. 00:00:23.995 --> 00:00:32.550 Well, first, our our series has a real focus on the intersection of mental health and technology. 00:00:33.410 --> 00:00:36.710 It's entitled humans over algorithms. 00:00:38.210 --> 00:00:40.850 And I am to make it personal to start off. 00:00:40.850 --> 00:00:45.975 I'm Brian Lohman, alumni coordinator here at Divine Mercy University. 00:00:46.275 --> 00:00:56.295 I've been here, about nine months and actually come from a technology background myself as a corporate recruiter within the government and commercial world. 00:00:56.900 --> 00:01:02.900 And I also am in the master's in counseling program. 00:01:02.900 --> 00:01:07.960 So I have an eye toward becoming a clinician down the road. 00:01:08.580 --> 00:01:12.875 For this series, we've already had a thousand participants thus far. 00:01:12.875 --> 00:01:14.175 We're only halfway through. 00:01:14.875 --> 00:01:20.095 We have, after Jerome this evening, we have May 28, Kathy Erwin. 00:01:20.875 --> 00:01:24.255 She is one of our, counseling, professors. 00:01:25.050 --> 00:01:28.750 And then, June 3, Garrett Boyer and Laura Cusimano. 00:01:29.450 --> 00:01:34.510 Doctor Boyer and Doctor Cusimano, they met here at DMU in the PsyD program. 00:01:34.970 --> 00:01:40.135 So, they're going to be doing a tag team as a husband and wife couple. 00:01:40.355 --> 00:01:41.975 So they are our last two. 00:01:43.315 --> 00:01:47.235 This evening, we're really grateful that Jerome is gonna be joining us. 00:01:47.475 --> 00:02:00.470 He's based in the San Francisco Bay Area and received his bachelor's in computer science at University of California, Riverside and his master's in psychology here at Divine Mercy University. 00:02:01.570 --> 00:02:13.495 I really think that the intersection of Jerome's expertise and his passion are a are a great fit for this midpoint webinar, just marrying technology and and faith. 00:02:14.355 --> 00:02:16.855 He's an engineering leader and engineering builder. 00:02:17.150 --> 00:02:23.490 He works really at that intersection of, artificial intelligence, human behavior, and real world systems. 00:02:24.190 --> 00:02:28.770 He's about a decade of expertise in, software engineering and leadership. 00:02:29.595 --> 00:02:40.095 He's led teams all the way from early stage startups to large scale organizations, building products that serve literally millions of users. 00:02:41.035 --> 00:02:56.310 He's also contributed to initiatives at the intersection of technology and the Catholic faith through collaborate excuse me, collaborations with Sacred Spark, Catholic Polytechnic University, TABELLA, and other mission driven organizations. 00:02:57.250 --> 00:03:07.545 His work really focuses on that realm of promoting trust, intentionality, and authentic human connection in an increasingly digital world. 00:03:08.485 --> 00:03:32.045 And as that technologist and a person of faith, he's really interested and has a passion for the ethical and psychological implications of AI, namely, you know, how systems influence attention, motivation, identity, and how we can really design technology that respects the the dignity of each and every human person. 00:03:32.045 --> 00:03:35.985 He's a great example of somebody who is in the world, but not of it. 00:03:36.125 --> 00:03:44.150 He's working in a technical corporate secular realm, but also with a foot in the the church with his work. 00:03:45.330 --> 00:04:03.805 And it's just a great witness to see him, a face of DMU out in the world, and we're really overjoyed to welcome Jerome as he shares his his expertise, his passion, and his faith, as he speaks on the theme of, augmenting care, not replacing it. 00:04:04.265 --> 00:04:11.085 In one technical note, there'll be plenty of questions, I'm sure, which, come to mind, that Jerome is happy to address. 00:04:11.640 --> 00:04:14.040 He and I chatted that we would like to do that throughout. 00:04:14.040 --> 00:04:27.695 So, if there's a point he makes that you want to clarify or you want to follow-up or you just would like to comment, please, pop a note in the chat box and I will be sharing that with Jerome so he can respond to it throughout. 00:04:28.555 --> 00:04:32.575 So with that, let's hand things over to Jerome Placido. 00:04:32.715 --> 00:04:33.995 And, welcome, Jerome. 00:04:33.995 --> 00:04:35.055 Great to have you. 00:04:35.275 --> 00:04:35.755 Thank you. 00:04:35.755 --> 00:04:36.395 Thank you, Brian. 00:04:36.395 --> 00:04:38.175 And, thank you, everyone, for joining. 00:04:39.760 --> 00:04:41.940 Super excited and so happy to be here. 00:04:42.160 --> 00:04:45.860 As as Brian mentioned, I'm I'm a proud alumnus of Divine Mercy University. 00:04:46.160 --> 00:04:48.820 In 2020, I completed my master's of science in psychology. 00:04:49.600 --> 00:04:57.295 And, you know, I've been in in technology for ten, fifteen years, working in software and engineering leadership. 00:04:57.295 --> 00:05:09.075 Those who know me know I'm always trying to marry and see the interesting intersections between the two things I'm very passionate about, which is technology and my Catholic faith. 00:05:09.750 --> 00:05:23.625 In more recent years, my work has focused a lot on AI systems, on operational scalability, on data quality, how we evaluate AI to ensure that it's safe, it's accurate, that we're doing the right things with it. 00:05:23.945 --> 00:05:28.285 And more practically, just using a generative AI. 00:05:28.905 --> 00:05:31.945 I have my own AI bot that I stood up as a weekend project. 00:05:31.945 --> 00:05:33.085 His name is Leo. 00:05:33.785 --> 00:05:42.760 He sits here on my desk, and he helps me to schedule meetings, answer emails, do research for webinars, things like that. 00:05:43.060 --> 00:05:49.960 In my current role, my my day job when I'm not AI Batman at night, my current role is at Turing. 00:05:50.115 --> 00:05:56.935 I work on initiatives related to improving quality and reliability on AI driven workflows and operational systems. 00:05:57.875 --> 00:06:01.475 Outside of that professional work, I experiment a lot. 00:06:01.475 --> 00:06:15.435 And I, as Brian mentioned, I I, work a lot with a lot of mission driven Catholic organizations, consulting and and helping them leverage these technologies in a in a in a meaningful way. 00:06:16.075 --> 00:06:23.615 I do want to note that today I am speaking independently, not as a in an official capacity for my employer or any affiliated organizations. 00:06:24.395 --> 00:06:26.975 These perspectives that I share and opinions are my own. 00:06:27.275 --> 00:06:42.200 And I also just want to take the time now because I've experienced so much growth, and I'm so grateful to have been a part of a cohort in Divine Mercy University that I just want to express my gratitude for being invited to be a part of this conversation today. 00:06:42.200 --> 00:06:45.285 So I look forward to the discussion that we're gonna have. 00:06:45.285 --> 00:06:49.545 And there's a lot I think that I'm going to bring up that is maybe a little spicy. 00:06:50.885 --> 00:06:53.145 But I definitely think it's worth talking about. 00:06:54.245 --> 00:06:55.385 So, yeah, absolutely. 00:06:55.445 --> 00:07:01.220 If there's any questions at all, like Brian had said, we'll just raise them in the chat and we'll answer them ad hoc. 00:07:02.160 --> 00:07:16.645 So before we begin the actual conversation of what the intersection is between psychology and and health care in general, I thought it might be good to ground us in some some concepts. 00:07:17.505 --> 00:07:19.125 So what is artificial intelligence? 00:07:19.185 --> 00:07:29.430 Well, AI or artificial intelligence is a broad field of computer science focused on creating systems capable of performing tasks that typically require human intelligence. 00:07:29.970 --> 00:07:37.430 Pattern recognition, decision making, these sort of tasks, when done by a computer, we consider it to be called artificial intelligence. 00:07:38.210 --> 00:07:42.455 Within this realm of artificial intelligence, there's some other terms that we should be aware of. 00:07:42.455 --> 00:07:53.810 And these are sort of building blocks to to help us understand what we know today in in the commercial realm realm is like chat g p t or Claude or any of those other commercial apps. 00:07:54.590 --> 00:07:56.450 First, machine learning. 00:07:56.990 --> 00:07:58.450 What is machine learning? 00:07:58.590 --> 00:08:08.645 Machine learning is a lot of classification and ranking, a lot of statistics and math to help create risk and prediction models across the dataset. 00:08:09.025 --> 00:08:16.965 So we are going to take a dataset and do pattern detection on those datasets to help predict other parts or other things about that data. 00:08:17.665 --> 00:08:20.065 That's machine learning in its traditional sense. 00:08:20.065 --> 00:08:26.340 It's systems that learn patterns to make data or to learn patterns from data to make decisions. 00:08:28.000 --> 00:08:32.100 Next, is generative AI, and this builds on machine learning. 00:08:32.160 --> 00:08:41.675 So instead of just predicting outcomes, a, b, c, and what's next is d, it also generates new content like text, images, code, or audio. 00:08:42.055 --> 00:08:45.435 It works by learning patterns from massive datasets. 00:08:45.975 --> 00:08:49.355 And these datasets are used to train the models. 00:08:49.655 --> 00:08:54.240 And then it will predict the most likely next piece of information. 00:08:56.140 --> 00:08:58.300 This is a very rudimentary way of thinking about it. 00:08:58.300 --> 00:09:06.585 But if you if anyone has an iPhone or it's also on Android as well, but, autocomplete, when you say, hello, how are you? 00:09:06.585 --> 00:09:10.925 And then at the bottom of that tab on on autocomplete, it says doing. 00:09:11.385 --> 00:09:15.965 It's predicting what your what your next thing that you're gonna say is. 00:09:16.185 --> 00:09:28.900 That is more likely in the machine learning realm, but it delves a little bit and helps us to to maybe see more concretely what we mean by the likely next piece of information. 00:09:29.360 --> 00:09:32.260 So we train these models on a ton of data. 00:09:32.845 --> 00:09:47.580 Initially, it was data across the entirety of of the Internet and then use that data to find these patterns to help predict what's the likely next word, sentence, phrase, or thing that should be generated within this model. 00:09:48.060 --> 00:09:49.680 Hence, generative AI. 00:09:50.700 --> 00:09:56.620 Now, conversational agents is probably what most of us have been have used or are starting to use. 00:09:56.620 --> 00:09:57.340 It's very funny. 00:09:57.340 --> 00:10:04.505 At my parish, we were preparing for a gala, a fundraising gala to help build our church. 00:10:04.965 --> 00:10:14.185 And I had everyone from folks who have been retired using chatGPT to college students using chatGPT. 00:10:14.325 --> 00:10:15.845 And so it's across the board. 00:10:15.845 --> 00:10:17.625 A lot of people are using these things. 00:10:17.720 --> 00:10:20.460 But what they're using really are conversational agents. 00:10:20.760 --> 00:10:34.785 Now conversational agents take the generative AI that has been developed, and it they create a system, and maybe let's consider it a harness to help these generative AI models do more. 00:10:34.785 --> 00:10:44.325 So it combines them with memory, with tools, with workflows and actions so that it's not just generating a response, but they can assist with tasks. 00:10:44.385 --> 00:10:53.020 So for example, when we go and chat gpt, and we say, you know, tell me everything you know about Pope Leo the fourteenth. 00:10:53.400 --> 00:11:03.215 It's not only gonna take the the information it was trained on, which if you remember maybe two, three years ago, was trained only up until 2022. 00:11:03.355 --> 00:11:06.495 And so there was a limit on its world of knowledge. 00:11:06.955 --> 00:11:12.155 Since then, they've added tools to enable it to search the web so I had more relevant information for you. 00:11:12.155 --> 00:11:17.780 They've created a harness around these models so that they are able to do more. 00:11:18.560 --> 00:11:20.340 Simple way to think about it is this. 00:11:20.480 --> 00:11:22.900 We have we have machine learning that learns. 00:11:22.960 --> 00:11:29.780 We have generative AI that creates and the conversational agents that assist and act the feet of the model. 00:11:31.875 --> 00:11:32.355 Great. 00:11:32.355 --> 00:11:38.755 So that's a I know it was a little little long, but I think these nuances help to dispel some of the ideas. 00:11:38.755 --> 00:11:52.160 As soon as you know that generative AI are are the fuel of conversational agents and that it's more a statistical model of predicting what's the likely next word that the user is going to expect. 00:11:53.020 --> 00:11:57.360 It's a lot harder to believe that these models are thinking. 00:11:58.895 --> 00:12:06.175 And so there's a lot of nuance around the actual complexity of those prediction machines, but that is exactly what they do. 00:12:06.175 --> 00:12:06.835 They predict. 00:12:07.775 --> 00:12:12.930 Now, let's frame artificial intelligence against the backdrop of psychology today. 00:12:13.410 --> 00:12:18.150 Most of you who have a background in psychology or part of the industry know this already. 00:12:18.210 --> 00:12:28.550 And I know there's other things to be addressed, but I just want to maybe address one of the pressures here that may lead to us using and potentially misusing artificial intelligence. 00:12:29.755 --> 00:12:34.495 Let's take a look at the math of the global mental health crisis as it exists today. 00:12:34.795 --> 00:12:44.255 Currently, in the world today, according to the World Health Organization in 2025, there are nine hundred and seventy million people worldwide that experienced a mental illness. 00:12:45.020 --> 00:12:53.840 That is if you stood people shoulder to shoulder around the world, we could wrap the world about 12 times. 00:12:56.620 --> 00:13:00.725 Now with that particular burden, the need, there's also a gap. 00:13:00.725 --> 00:13:04.745 There just aren't enough providers to serve for those nine hundred and seventy million. 00:13:05.045 --> 00:13:07.705 Some of those because of disparity in their location. 00:13:08.005 --> 00:13:09.945 Some of those because of lack of access. 00:13:10.660 --> 00:13:14.280 But just in general, even where there is some access, it's still hard. 00:13:14.500 --> 00:13:21.240 The economic cost to that, according to The Lancet, is $5,000,000,000,000 in global economic losses annually. 00:13:22.260 --> 00:13:39.225 This the statistic, while it it shows the the large impact from a from a maybe economic standpoint, doesn't also account for the emotional toll of the suffering of nine hundred and seventy million and the lack of ability that we can provide to help them. 00:13:39.640 --> 00:13:41.340 And there's a real bottleneck here. 00:13:41.480 --> 00:13:48.220 In The US specifically, twenty three percent of individuals will wait over twelve weeks for a face to face care. 00:13:48.760 --> 00:13:53.900 And the longer waits, as we know, actively exacerbate patient conditions. 00:13:56.285 --> 00:13:59.585 This is part of the landscape that we live in today. 00:14:00.045 --> 00:14:07.325 And so and by the way, this particular crisis, you can frame it with other parts in other industries. 00:14:07.325 --> 00:14:10.305 And so you have this overwhelming need of something. 00:14:12.180 --> 00:14:18.600 And so AI seems really seductive and tempting at this point because AI doesn't sleep. 00:14:18.660 --> 00:14:20.100 It's available twenty four seven. 00:14:20.100 --> 00:14:24.280 You can access it twenty four seven, three in the afternoon, three in the morning. 00:14:24.725 --> 00:14:27.785 It eliminates geographic and temporal barriers. 00:14:28.005 --> 00:14:33.785 It doesn't matter whether I'm California or I'm in London or I'm in Bangalore, India. 00:14:34.405 --> 00:14:39.865 And it's scalable to help support isolated or marginalized populations. 00:14:41.070 --> 00:14:54.530 In addition to that, there have been certain studies that have shown that people who have used some of these commercial chatbots have had a high sense of zero judgment zones. 00:14:55.035 --> 00:15:00.255 This is mainly because in speaking with a computer, there's less apprehension of sharing things. 00:15:00.315 --> 00:15:04.655 And so there's a reduced reduced social anxiety of impression management. 00:15:05.835 --> 00:15:17.860 And then in addition to that, you theoretically could scale this to millions of users given the right infrastructure, given the availability of of the, you know, compute. 00:15:18.400 --> 00:15:22.180 You could theoretically scale this to address the gap. 00:15:22.885 --> 00:15:27.465 And there's a near zero marginal cost for the information delivery. 00:15:28.485 --> 00:15:32.005 Now all of this makes it really attractive to say, hey. 00:15:32.005 --> 00:15:43.010 Let's use AI, which says these really nice and pretty things to answer and address this gap. 00:15:43.310 --> 00:15:50.290 Now, unfortunately, this gap exists today, and this is the world that we live in. 00:15:50.670 --> 00:15:58.465 And there are already individuals today who are using commercial chatbots as a quote emotional sanctuary. 00:15:58.465 --> 00:16:06.410 And that is the term they have used in a study done in 2024 to describe their interactions with AI chatbots. 00:16:06.630 --> 00:16:09.370 So this isn't this isn't a theoretical. 00:16:09.750 --> 00:16:31.315 People are going to these chatbots at today, to seek support, empathy, care, and even to triage difficult situations, like breakups with a boyfriend or girlfriend, estranged relationships with fathers, and finding a certain sense of comfort in these interactions. 00:16:32.720 --> 00:16:35.540 So that is sort of where we're at today. 00:16:36.800 --> 00:16:51.915 Now, one of the things I love about our faith in the Catholic churches is that the timeless teachings of the church have have been able to been projected against the idea of artificial intelligence. 00:16:52.455 --> 00:16:54.955 But we're definitely deepening in that understanding. 00:16:55.095 --> 00:17:00.920 And Pope Leo has definitely been leading the way, especially since he entered into his pontificate. 00:17:01.860 --> 00:17:22.885 I actually used AI just a couple nights ago, and I asked it, I asked Leo, my my chatbot, to do a deep research and pull all of the resources of all the different times Pope Leo has mentioned AI in any any particular address, whether it's a address for World Communications Day or address to the participants of any jubilee of governments. 00:17:23.105 --> 00:17:28.500 A lot of these things and it pulled back a ton and he has spoke a ton about AI. 00:17:29.040 --> 00:17:34.100 One of the things that he said, and he has acknowledged that AI is not going anywhere. 00:17:36.240 --> 00:17:42.835 And it's not that it's a technological problem, but more so that it's an anthropological problem. 00:17:44.975 --> 00:18:04.590 And what he means by that is that our understanding of the human person needs to needs to be firmed firmed up and and and bolstered so that we can approach the use of this technology in a certain way, especially when you have a technology that makes it sound like it's being empathetic towards you. 00:18:05.530 --> 00:18:15.665 The simulation it encroaches on human relationship death And then human memory is, he says, is generative and machines are static. 00:18:15.665 --> 00:18:20.885 And he's just continues to try to remind us sort of the difference between these two. 00:18:21.505 --> 00:18:27.500 He also said to a group of young people to avoid digital bubbles that weaken discernment. 00:18:30.680 --> 00:18:48.675 And more importantly for this for this group here, he's spoken specifically to folks in health care and reminds us that care is a noble proportion is noble in proportion to fragility, and that we should maintain the patient provider relationship. 00:18:49.935 --> 00:18:56.660 And then he urged us that AI should enhance but never replace the nobility of that care. 00:18:58.080 --> 00:19:00.000 Now, he's it. 00:19:00.000 --> 00:19:11.965 Unfortunately, the market and the conditions that that AI is being developed in really make it hard to create an ethical sort of scaffolding around AI. 00:19:12.265 --> 00:19:25.620 But nonetheless, as most of you know, he's definitely trying to tackle that, especially with his upcoming encyclical, I believe, magnifica humanitas, or magnus magnificent humanity. 00:19:26.160 --> 00:19:28.660 I'm really looking forward to the the encyclical. 00:19:28.880 --> 00:19:36.580 I was disappointed that it was delayed because I was hoping to just comb through it and read it 10 times over in preparation for this webinar. 00:19:36.880 --> 00:19:42.895 But maybe Brian, if you'll have me again, maybe we can discuss it further in a future webinar. 00:19:43.595 --> 00:19:45.515 But he's definitely trying to address that. 00:19:45.515 --> 00:20:09.365 And I think it's important because the market and the the ecosystem that AI is being created in today, its incentives for the creation and furthering of AI are more for capital gains to increase the the profits of shareholders and to increase the value of those particular private organizations. 00:20:10.385 --> 00:20:32.030 And so when those become the sole, motivators within these organizations, or maybe not the sole, but the strongest motivators, we we we tend to, as Pope Leo has said, as Pope Francis has said, we tend to push to the side the marginalized, those who are forgotten and the weakest in our society. 00:20:32.490 --> 00:20:43.225 And so Poplio calls for a global governance and something to help ensure the common good equity and reliability of these technologies. 00:20:48.850 --> 00:20:49.010 Great. 00:20:49.010 --> 00:21:04.315 Now I'd like to spend a couple minutes as we talk about just AI as a whole to talk a little bit about, some of the some of the potholes that that we may encounter in using it. 00:21:04.935 --> 00:21:17.360 I should I should mention that this the thesis and may may not be apparent in my introduction, but the thesis is that we can and should use AI, but there are definite things that we should be aware of. 00:21:17.420 --> 00:21:30.545 And part of my goal in this presentation is to help you to be aware of these potholes, these caveats, so that when you do use it, you're aware of when you have to dive deeper or maybe look for a different solution. 00:21:32.605 --> 00:21:36.225 Jerome, if I could just pop in with a question on my own Sure. 00:21:36.525 --> 00:22:02.555 To address to the the audience is, have you found, as it is with many things, not just, new technologies or, new challenges or opportunities, but, I think just within human nature that the extremes are always easier to embrace than sort of the the the via media, the middle way of salvation, if you will. 00:22:02.555 --> 00:22:07.535 So in this realm, it's it's an easy thing to say, well, listen. 00:22:07.675 --> 00:22:08.335 You know? 00:22:08.510 --> 00:22:09.010 No. 00:22:09.070 --> 00:22:17.970 AI isn't from God, but, you know, it comes from, the the hearts and minds and hands of of those who were given gifts by God. 00:22:18.750 --> 00:22:23.845 And so, it it just needs to be There's no concerns in it. 00:22:23.845 --> 00:22:25.525 It's just how it's ethically used. 00:22:25.525 --> 00:22:27.925 There's nothing baked negatively into it. 00:22:27.925 --> 00:22:29.625 But I think of the Sure. 00:22:29.645 --> 00:22:34.165 Some of the engineers from the social media platforms who said, No, no, no. 00:22:34.165 --> 00:22:36.425 It's not just that these things are used nefariously. 00:22:37.170 --> 00:22:38.930 It was baked in to be addictive. 00:22:38.930 --> 00:22:40.290 Like, that's how we made it. 00:22:40.290 --> 00:22:46.070 So there's one extreme, which is just, well, you know, it's always gonna be good. 00:22:46.690 --> 00:22:48.850 There's nothing negative inherently about it. 00:22:48.850 --> 00:23:01.715 Or the other extreme to say, well, we we rejected wholeheartedly, which means that the people like you and others are not gonna be guiding us on it. 00:23:01.715 --> 00:23:13.910 And so, do you see people sort of tending to one extreme or another versus sort of that middle way of, listen, there's guards around it, Leo's writing on it, and I'm grateful he is. 00:23:14.610 --> 00:23:19.350 Do you find that we're just sort of trying to get our get our bearings and find that middle way? 00:23:19.645 --> 00:23:22.925 You know, I think I think getting your bearings is the right way to really look at this. 00:23:22.925 --> 00:23:25.725 I think people don't know what they don't know yet. 00:23:25.725 --> 00:23:41.420 And while, for a technologist like me, I mean, the second that ChatGPT came out in 2020, '20 or yeah, around 2020, I popped it up and I had it compose a prayer. 00:23:41.560 --> 00:23:43.400 I'm like, hey, compose a Catholic prayer for me. 00:23:43.400 --> 00:23:44.540 And it did it beautifully. 00:23:45.195 --> 00:23:49.535 I could even argue that maybe that was the first Catholic thing it did in the entirety of GPT. 00:23:50.715 --> 00:23:57.135 But it's even though it's that old, five years old, it's still fairly new to a lot of the population. 00:23:57.275 --> 00:24:06.350 So it's more of a wait and see moment for for maybe people who aren't used to using these technologies or don't know enough about it. 00:24:06.410 --> 00:24:15.545 That being said, though, Pope Leo, in an address has mentioned that this technology is not neutral. 00:24:15.545 --> 00:24:18.445 We can't we can't just sit here and think that it is neutral. 00:24:19.945 --> 00:24:34.440 Not because there are nefarious things behind it, but because there is a worldview that has gone into either the development of this technology or even the the, the data that's that, the models are trained on. 00:24:34.440 --> 00:24:46.255 And that's part of what I I plan on presenting here is that, it's not so much that it's not so much that there are sort of like, there is a middle way. 00:24:46.255 --> 00:24:49.135 I think we're trying to navigate to find what that middle way is. 00:24:49.135 --> 00:24:56.810 But I think that there are a lot of caveats that people may not be aware of to help them navigate to that middle way. 00:24:56.810 --> 00:24:57.870 Does that make sense, Brian? 00:24:59.770 --> 00:25:00.970 That's perfect sense. 00:25:00.970 --> 00:25:01.470 Exactly. 00:25:01.850 --> 00:25:03.210 Thanks for responding to that. 00:25:03.210 --> 00:25:03.530 Yes. 00:25:03.530 --> 00:25:05.550 Anybody else who wants to jump in, please. 00:25:06.570 --> 00:25:07.050 Awesome. 00:25:07.050 --> 00:25:07.790 Thank you. 00:25:07.850 --> 00:25:08.670 Thanks, Jerome. 00:25:08.730 --> 00:25:08.970 Yeah. 00:25:08.970 --> 00:25:09.470 Absolutely. 00:25:09.530 --> 00:25:24.565 So getting into those caveats, one of the hard things about the use of artificial intelligence in general and large language models is it essentially is a black box problem. 00:25:25.440 --> 00:25:31.060 Even for myself who uses it as a developer, you sort of poke at it. 00:25:31.360 --> 00:25:32.260 It says something. 00:25:33.520 --> 00:25:34.560 You poke at it again. 00:25:34.560 --> 00:25:37.360 It could say something completely different the next time around. 00:25:37.360 --> 00:25:39.435 It's not deterministic at all. 00:25:39.435 --> 00:25:48.075 And the decision logic that it makes or that it uses to say different things are really hidden because of the way that it's been developed. 00:25:48.075 --> 00:25:49.135 It's very convoluted. 00:25:49.275 --> 00:25:56.680 The deep learning models make it impossible to trace the exact rationale behind some of the things that it says. 00:25:58.100 --> 00:26:06.760 That in turn used in a psychiatric or mental health setting makes it really hard to understand why something is said. 00:26:07.620 --> 00:26:12.565 There's also a deficit in validation sometimes. 00:26:12.865 --> 00:26:26.350 Now, there are extensive quality assurance processes that both the these organizations have in place to check that a model is saying the things within its boundaries. 00:26:26.650 --> 00:26:32.830 And even some quality, there's a lot of quality checks even for the the training data that's going into these models now. 00:26:32.890 --> 00:26:51.785 So but while those are there, it's impossible in the real world to really draw the right boundaries to make sure that it is not making, you know, irrelevant correlations or saying the wrong thing at the wrong time. 00:26:54.260 --> 00:26:59.880 There's also a lack of of, or ambiguity and liability. 00:27:00.420 --> 00:27:05.960 If something happens and we're using a chatbot in in in the mental health setting, who's liable? 00:27:06.575 --> 00:27:08.335 Is it the developer who wrote the code? 00:27:08.335 --> 00:27:09.315 Is it the CEO? 00:27:09.775 --> 00:27:17.795 There's aren't any there aren't any things in place to understand when something does go wrong, who's to blame? 00:27:18.095 --> 00:27:23.910 Now, I do wanna I do wanna share a recording. 00:27:23.970 --> 00:27:35.490 One of the things I did to prepare for for this webinar today was and something that I do quite frequently is that I will take my thesis, my premise of whatever I'm working on this. 00:27:35.490 --> 00:27:36.375 I do this at work. 00:27:36.455 --> 00:27:38.715 I do this in preparation for talks. 00:27:39.015 --> 00:27:46.635 And I and I tell AI to sort of present it back to me and try to poke holes at it and sort of play devil's advocate. 00:27:47.735 --> 00:27:52.290 However, this time, instead of doing that, I just gave it all the resources that I was reading that I was combing through. 00:27:52.290 --> 00:27:59.750 And I'm like, you know, I just wanna go for a run, and I'm gonna have it synthesize the data and just read it back to me in whatever way that it comes back. 00:27:59.810 --> 00:28:06.595 And I'm gonna share that that recording now. 00:28:06.975 --> 00:28:08.355 And this is, unprompted. 00:28:08.575 --> 00:28:16.355 This is just it telling me sort of the the, synthesized, version of the sources that I gave it. 00:28:17.060 --> 00:28:22.820 Imagine pouring your deepest, darkest trauma out to a therapist who never ever judges you. 00:28:22.820 --> 00:28:23.320 Right. 00:28:23.380 --> 00:28:29.395 A companion who, you know, perfectly tailors their personality to your exact emotional needs. 00:28:29.395 --> 00:28:31.875 And who never gets tired of hearing you loop through your anxiety. 00:28:31.875 --> 00:28:32.195 Yeah. 00:28:32.195 --> 00:28:32.675 Exactly. 00:28:32.675 --> 00:28:35.955 And who is available to talk at literally three in the morning. 00:28:35.955 --> 00:28:38.195 Which is, I mean, unheard of in traditional therapy. 00:28:38.195 --> 00:28:38.695 Totally. 00:28:38.835 --> 00:28:46.480 Now imagine finding out that this profound, life changing, emotional sanctuary is actually just a string of code. 00:28:49.020 --> 00:28:49.840 That's it. 00:28:50.060 --> 00:28:56.320 Now for maybe the, untrained, ear, that sounds fine. 00:28:56.860 --> 00:28:59.455 I hear a little bit of bias there. 00:28:59.515 --> 00:29:20.020 And me knowing the me knowing the data that I put into it, which was about 60%, Pope Leo cautioning against the use of data, 40% neutral and actually some sort of like naysayers of not using data in the field of psychology. 00:29:20.560 --> 00:29:26.765 The fact that AI took that and gave me this output, I can't tell you why it did that. 00:29:26.825 --> 00:29:39.405 But I can tell you that it it's it seems really apparent in this particular recording that it was for the use of this and it actually said at the end it wrapped up in this this podcast that I had to create. 00:29:39.420 --> 00:29:57.385 If an AI can perfectly mimic a loving and insightful therapist, and as a direct result of that interaction, your brain genuinely releases serotonin, your physical anxiety drops, and your depression actually fades, does it truly matter that the AI doesn't actually feel empathy? 00:29:59.765 --> 00:30:09.330 And so now granted, none of this this is sort of just posing the question and and it's it could be considered neutral. 00:30:09.470 --> 00:30:19.490 But I in this was surprised to sense a little bit of swaying in the direction of one way versus the other way with just the neutral resources that I gave it. 00:30:19.975 --> 00:30:21.115 This is just an example. 00:30:21.255 --> 00:30:23.095 I'm not saying that this has happened all the time. 00:30:23.095 --> 00:30:26.135 I ran it again against the same set. 00:30:26.135 --> 00:30:30.395 I had it create another another podcast and it was completely different. 00:30:30.535 --> 00:30:32.875 It was more neutral and more balanced. 00:30:33.840 --> 00:30:40.660 So just an example of of sort of the opaque decision logic it uses to synthesize data for us. 00:30:42.960 --> 00:30:45.620 Before you go, we have a question from Violetta. 00:30:45.760 --> 00:30:46.960 Give me a second. 00:30:46.960 --> 00:30:51.885 Let me Violetta, you had had a, a point you want to make? 00:30:54.425 --> 00:30:55.245 Hey, Violetta. 00:30:55.465 --> 00:31:00.685 I I would like to, ask if, this, we will have access. 00:31:00.745 --> 00:31:13.400 Like, I am a middle school and high school teacher, and I would love to be able to, share this with, my students. 00:31:15.075 --> 00:31:16.435 We will be yes. 00:31:16.595 --> 00:31:20.515 We are recording, and we'll be sharing this on, DMU's YouTube channel. 00:31:20.515 --> 00:31:23.395 So, yes, please share this Y Beyond and with your students. 00:31:23.395 --> 00:31:29.495 And and, what what, something is speaking to you about wanting to share this with them? 00:31:30.780 --> 00:31:31.420 I'm sorry. 00:31:31.420 --> 00:31:38.240 I I couldn't, There's some relevant points that Jerome is, discussing you'd like to share with your students? 00:31:39.100 --> 00:31:39.500 Yes. 00:31:39.500 --> 00:31:42.160 I think that, the information is very valuable. 00:31:42.300 --> 00:32:33.640 And, like, the instructor is is curious about, you know, that I, being in the that I am in the classroom, they, well, the students right now, like, they are My point is that the, the district is very, concerned about how they are training us to, to be able to, teach the students, but at the same time as, I forgot the person that is presenting, it's state stating, it is very hard to compete with AI because this is open twenty four seven no matter what geographical part of the world you are. 00:32:34.020 --> 00:33:01.580 And then it's it's like, how can you, as a teacher, you can teach and not, impart it to your students that, you know, to use their in intelligence to because right now, like, if you give them, for example, an essay or even asking a question on the classroom, they go they go to JWT to do anything. 00:33:01.640 --> 00:33:04.200 They are not actually thinking. 00:33:04.200 --> 00:33:19.495 And when he was sharing about, you know, that how this, intelligence is, like, it's not going anywhere, but how can they are able to protect themselves against themselves? 00:33:20.195 --> 00:33:22.135 Because it's like they are cheating themselves. 00:33:23.110 --> 00:33:23.610 Sure. 00:33:24.150 --> 00:33:24.650 Yeah. 00:33:24.710 --> 00:33:26.650 Violet, I think that you bring up a valid point. 00:33:27.190 --> 00:33:40.305 I I do think, part of my background, by the way, I did about ten years in in education technology, and my most recent experience was, helping to sort of implement AI systems within an EdTech system. 00:33:41.005 --> 00:33:48.545 And so for someone from the outside looking in to teachers, I my heart goes out to you. 00:33:49.085 --> 00:33:52.545 And to navigate all this, I think is really, really hard. 00:33:52.830 --> 00:33:56.690 I have been passionate about EdTech for such a long time and education as a whole. 00:33:56.830 --> 00:34:11.575 I do think that some of similar to what we need to do here in psychology, some of the ways that we teach, some of the methodologies we use fundamentally have to change and evolve because AI is not going anywhere. 00:34:12.275 --> 00:34:34.700 And I think that the challenge is on now administrators, teachers to make sure that one, these systems are safe, and two, that they continue to be aware of these caveats while also promoting the things that we want to capture within tech education itself, which is human flourishing and the growth of that that individual, both emotionally but also intellectually. 00:34:35.075 --> 00:34:38.195 And so, yeah, I completely agree with you. 00:34:38.355 --> 00:34:40.515 Part of me actually is eager. 00:34:40.515 --> 00:34:41.715 I was telling Brian this earlier. 00:34:42.355 --> 00:34:47.175 I'd love to be a student again, just to see what it's like to be a student in this AI world. 00:34:47.810 --> 00:34:58.870 Because it is a lot easier now, for example, to study a particular subject if you have the right tools and methodologies to employ know what right tools and methodologies to employ. 00:34:59.425 --> 00:34:59.925 Right? 00:35:00.465 --> 00:35:14.565 Similarly, for those in psychology, when the next 20 research papers come out, it's a lot easier to intake that information now using artificial intelligence. 00:35:16.200 --> 00:35:16.700 Great. 00:35:17.080 --> 00:35:17.960 We had one more. 00:35:18.280 --> 00:35:19.720 I'm not sure if you wanna make a comment. 00:35:19.720 --> 00:35:21.320 It's because we have a few teachers here tonight. 00:35:21.320 --> 00:35:25.160 Adriana, you were making a point that, yes, you you are in the classroom as well. 00:35:25.160 --> 00:35:26.940 You you understand the concerns. 00:35:28.360 --> 00:35:30.310 Did you have something you wanted to offer? 00:35:41.355 --> 00:35:42.155 Sorry, Adriana. 00:35:42.155 --> 00:35:43.035 I needed to unmute you. 00:35:43.035 --> 00:35:44.315 You can go ahead if you'd like to chat. 00:35:44.315 --> 00:35:44.815 Hello? 00:35:45.130 --> 00:35:45.630 Yes. 00:35:45.930 --> 00:35:46.750 Oh, well, yeah. 00:35:47.050 --> 00:35:50.990 Really concerned because I work with neurodivergent students. 00:35:51.370 --> 00:35:51.770 Mhmm. 00:35:51.770 --> 00:35:58.510 And we keep telling them that they are, you know, as important as anyone else. 00:35:58.970 --> 00:36:10.985 But how can they even compete when this is going just, you know, so so fast? 00:36:10.985 --> 00:36:13.405 And with them, it's like baby steps. 00:36:13.785 --> 00:36:24.090 So just basically, oh, it it's I could hear the the, the tension in the in the teacher's voice. 00:36:24.090 --> 00:36:34.355 And, yeah, in a in in a moment, I panic, but I also remember that god is bigger, and he loves us all, and he has a purpose, and he has a plan. 00:36:34.495 --> 00:36:34.815 Yeah. 00:36:34.815 --> 00:36:38.035 But it's it's just kind of difficult not get getting overwhelmed. 00:36:38.255 --> 00:36:38.995 Thank you. 00:36:40.740 --> 00:36:41.060 Yeah. 00:36:41.060 --> 00:36:42.020 Thank you, Adriana. 00:36:42.020 --> 00:36:42.420 Yeah. 00:36:42.420 --> 00:36:42.900 I agree. 00:36:42.900 --> 00:36:44.260 It's a lot to keep up with. 00:36:44.260 --> 00:36:45.320 It's it is overwhelming. 00:36:46.500 --> 00:36:51.880 I do think, though, that, it also is a it's a matter of framing as well. 00:36:53.225 --> 00:37:03.645 For neurodivergent students or students maybe who need more attention and care, and this also translates to any helping profession, those under our care. 00:37:04.185 --> 00:37:17.630 If done correctly, it frees the the caretaker, the administrator, whoever it frees us up to pay more attention to them if we do it correctly. 00:37:17.630 --> 00:37:24.295 And so what I'd love to do is switch because I recognize that the maybe the first half of this was sort of like, oh AI. 00:37:24.295 --> 00:37:27.195 It's this really big thing and it's constantly changing. 00:37:27.415 --> 00:37:47.310 It changed, you know, this month versus six months ago, but there are things that we can do proactively today to really allow it to be the tool that helps us and accelerates us so that we can not only be more productive, but also just to be more free to do things that help us to flourish as human beings. 00:37:47.450 --> 00:38:16.580 Because I guarantee you that while it there might be professionals at it, there's probably more flourishing found outside of spending eight hours working in spreadsheets and making sure all the numbers add up and little things like these are things that AI can do really well that are low in in personal relational cost and high and just like this very mundane sort of task management stuff. 00:38:16.800 --> 00:38:27.165 Before I move there, though, I just do also wanna wanna highlight a couple other systemic risks that we do want to be aware of when we do start using AI or if we do. 00:38:27.225 --> 00:38:40.850 One is the emotional dependency making sure our this these emotional attachments sometimes when when people are using AI lead to social isolation becoming overly dependent on on these bots. 00:38:40.910 --> 00:38:42.690 Another one is an automation bias. 00:38:43.390 --> 00:38:53.835 This is something that I have to watch out for myself or anyone using AI just in the same way that twenty years ago, I knew everyone's phone numbers back of my head. 00:38:53.835 --> 00:38:57.515 You could you could ask me for a phone number and I would I would give it to you. 00:38:57.515 --> 00:38:59.835 I only know two phone numbers now. 00:38:59.835 --> 00:39:00.510 That's it. 00:39:00.590 --> 00:39:05.650 I don't know any other phone numbers because my mind is not exercising that particular muscle anymore. 00:39:05.870 --> 00:39:12.430 Now it's a lot easier to fall into that trap with automation bias and having LLMs do a lot of these things for you. 00:39:12.430 --> 00:39:25.735 So it's very important to select the right things to offload to a, to an AI agent or an AI or a LLM model and to take on the other part for yourself. 00:39:30.550 --> 00:39:34.870 So how does this look in practice then if we do wanna use AI? 00:39:34.870 --> 00:39:43.485 Because, again, my position here is that we should and it would be it would be hard not to in the years moving forward. 00:39:43.485 --> 00:39:47.085 A lot of the things that we're doing now, they're already whether or not you like it. 00:39:47.085 --> 00:39:50.305 They're already augmented even internally with artificial intelligence. 00:39:51.085 --> 00:40:08.440 So we can divide the labor of care with things that we can do algorithmically, which means things that have a pattern and a series of steps to it that need some sort of monitoring, continuous availability, and some standard processor extraction. 00:40:16.645 --> 00:40:17.865 First side is the system. 00:40:17.925 --> 00:40:22.745 The second side is the part that's more human, the thing that's more accountable morally and ethically. 00:40:24.050 --> 00:40:37.110 The person or the part of the system, the person, that is able to make the complex clinical judgment, and interpretation of nonverbal and subtle cues. 00:40:38.505 --> 00:40:55.040 I heard it really put really well earlier as I was listening to a couple of talks around this is, as I mentioned before about AI and large language models, AI is really good at noticing patterns behind words. 00:40:55.040 --> 00:40:57.360 As I mentioned earlier, it's a statistical model. 00:40:57.360 --> 00:41:06.825 So when it when you have a sentence, it can it can take that sentence, do some really cool math and determine what's the next word. 00:41:07.065 --> 00:41:13.085 What's the what's the closest thing that this group of words is like. 00:41:13.305 --> 00:41:16.445 It's good at analyzing the patterns of words. 00:41:17.065 --> 00:41:21.565 But humans were good at understanding the meaning behind words. 00:41:22.050 --> 00:41:24.470 So I'm not saying that we should not have AI. 00:41:24.610 --> 00:41:27.490 I'm saying let's let AI do what it's really good at. 00:41:27.490 --> 00:41:52.180 There's been some great studies that have helped us to understand that when a person is using firsts in describing, I forget the particular study here, but first person sort of descriptions of a situation, when done in a particular percentage or an excessive way, is a leading indicator of depression or anxiety. 00:41:52.960 --> 00:42:08.415 Now you for someone who might be a practitioner, a therapist sitting and trying to listen and and be empathetic and to maintain that therapeutic relationship, you may not notice that particular pattern while trying to mirror and empathize. 00:42:09.355 --> 00:42:14.815 AI is absolutely the right tool to help to pick up on some of those cues. 00:42:15.515 --> 00:42:21.610 The practitioner themselves should be the decision maker of how to react to that particular data. 00:42:21.610 --> 00:42:29.150 They still have the moral and ethical responsibility and accountability in that particular in that particular scenario. 00:42:30.490 --> 00:42:43.165 And in addition to all that, from my studies at Divine Mercy University and most people who know it and who've had good therapy sessions, know that part of the actual healing process is that therapeutic alliance. 00:42:43.165 --> 00:42:46.420 It's that relationship between yourself and the therapist. 00:42:46.420 --> 00:42:49.620 I often sometimes joke that finding a good therapist is like dating. 00:42:49.620 --> 00:42:56.120 It's really like making sure you have the right fit and finding the right connection with that particular individual. 00:42:56.580 --> 00:43:01.535 And so that emotional and personal connection is still there. 00:43:01.535 --> 00:43:02.675 It's still very important. 00:43:04.895 --> 00:43:19.750 And that connection enables the practitioner that even the teachers to understand the meaning behind the words, But we're still able to use systems like this to maybe look further and look at other maybe leading indicators for other issues. 00:43:22.930 --> 00:43:25.190 So what does that look like practically? 00:43:25.375 --> 00:43:25.695 Gosh. 00:43:25.695 --> 00:43:27.955 I am going a lot slower than I anticipated. 00:43:28.095 --> 00:43:29.875 So let me just go through this one here. 00:43:30.415 --> 00:43:42.000 For for helping professionals and practitioners, this looks really easy, but it's actually pretty hard to accomplish by yourself. 00:43:43.100 --> 00:43:52.000 Using AI for screening and intake support, for triage and risk flagging, for symptom monitoring and documentation, for administrative stuff especially. 00:43:52.460 --> 00:44:03.535 If you have to imagine having to reschedule and shuffle yours your your calendar and then sending emails to those respective individuals to explain the change. 00:44:03.755 --> 00:44:12.630 All of that, I'm sure takes time and administrative and most people who is don't have that skill or that desire to do these administrative things don't like to do that. 00:44:12.690 --> 00:44:13.990 AI does that really well. 00:44:14.210 --> 00:44:17.330 And also low intensity support tools, but in limited context. 00:44:17.330 --> 00:44:19.990 So chatbots and things but in very limited context. 00:44:20.985 --> 00:44:25.465 All these things, I would propose, are useful to a certain extent. 00:44:25.465 --> 00:44:29.385 Now, obviously, the screening and intake support, if you just say, hey. 00:44:29.385 --> 00:44:30.585 You wanna see me? 00:44:30.585 --> 00:44:33.990 Go talk to my chatbot, and then I'll see you after you're done with that. 00:44:34.070 --> 00:44:37.670 That's not an ideal experience, first of all. 00:44:37.670 --> 00:44:42.410 But second, it takes away the human relationship building that's probably there. 00:44:42.550 --> 00:44:58.725 But it could help, for example, if you're in taking in taking someone, if there is a common template that you normally fill out for a particular patient and using AI to take the transcript of a meeting and move it over into that particular intake form. 00:44:59.105 --> 00:45:00.085 Simple and easy. 00:45:01.100 --> 00:45:05.280 Obviously, there's implications with that and and, what's allowed and what's not allowed. 00:45:05.340 --> 00:45:09.920 I'm not taking that into account here, but there are systems that have been approved for for things like that. 00:45:11.980 --> 00:45:19.965 So the care workflow is that AI really is in the background and that anyone who's using AI should be in the foreground. 00:45:20.585 --> 00:45:27.805 AI now becomes the supporting structure that helps us to do a lot of our administrative tasks. 00:45:27.865 --> 00:45:30.285 It should help us for asynchronous monitoring. 00:45:31.020 --> 00:45:40.220 An example of this that I thought of now this is maybe not been done yet. 00:45:40.220 --> 00:45:48.705 It's an idea that I just had as I was thinking about this was what if you could take all the notes that you had of a client maybe that you've been seeing for ten years? 00:45:49.405 --> 00:46:04.410 And maybe with AI to leverage its capabilities to understand patterns, cycles, and triggers for those cycles so that you're more aware in a particular session and more present for those particular triggers. 00:46:04.410 --> 00:46:08.270 And maybe some of those triggers you're already aware of, but maybe some might be a surprise to you. 00:46:09.215 --> 00:46:16.035 That's just one thing one way to sort of asynchronously modern monitor someone in your care. 00:46:16.335 --> 00:46:29.460 But the top most important layer is that a human clinician, a human person interprets the data, provides the empathetic relational care, and exercises the judgment. 00:46:29.680 --> 00:46:32.720 It's always going to be a human in the loop call. 00:46:32.720 --> 00:46:38.825 We never take the human out of these things because one, why? 00:46:39.525 --> 00:46:46.585 Two, the part of the healing process or part of the the support process is that relational aspect. 00:46:47.045 --> 00:46:51.145 Now, as I mentioned before, I did attend the Divine Mercy University. 00:46:51.880 --> 00:46:57.740 The Catholic Christian meta model of the person has greatly impacted how I even build software. 00:46:58.520 --> 00:47:04.460 But it's also impacted how I interpret my human my understanding of the human person and the anthropology. 00:47:05.285 --> 00:47:25.730 And so I was as I was thinking about these things, I sort of use that, that framework to sort of create a chart here around what where we should have maybe AI and where AI is possible, where relational complexity is low, the patient vulnerability is low. 00:47:25.790 --> 00:47:28.770 We can probably have AI help with some of those things. 00:47:29.710 --> 00:47:40.025 Where relational complexity is high and patient vulnerability is high in those one on one sessions, triaging hard moments, a human is necessary there. 00:47:40.025 --> 00:47:48.420 And there's no space for AI to take any any part of the decision making process. 00:47:48.420 --> 00:47:53.700 Now you can obviously still use it to augment some of your some of your triaging and things like that. 00:47:53.700 --> 00:48:00.835 But in general, I I would follow something like this to help augment care while leveraging artificial intelligence. 00:48:05.295 --> 00:48:22.940 And so you have the intersection of these two circles where you have a system that supports the things that you do, And you have human tasks that are solely part of your realm and responsible for and accountable for. 00:48:23.080 --> 00:48:35.745 So, for example, the AI can understand the structure and content or context of of health records of socio demographics, and it can process massive amounts of data. 00:48:36.845 --> 00:48:47.250 And then the human clinician is more focused on connecting, judging, when to act, and responding in those empathetic ways to alleviate suffering. 00:48:48.910 --> 00:48:58.635 Ultimately, the real question is the kind of care and what kind of person our systems are forming. 00:48:59.015 --> 00:49:02.235 And we have to keep that in mind when we use artificial intelligence. 00:49:02.695 --> 00:49:27.935 If we just threw an AI chatbot at our patients, at those who are helping and saying, hey, use this, We have to understand sort of the biases that we, the biases maybe that the model has, the limitations and AI has and the sort of outcome and the system that you are creating in that particular environment. 00:49:28.155 --> 00:49:44.460 However, if you are bolstering your ability as a clinician to care for those in your care by increasing insights, making it easier to to take time with you. 00:49:44.460 --> 00:49:55.505 In my view, these particular technologies should multiply our ability to help more people, but not without a human in the loop. 00:49:56.125 --> 00:49:57.405 A human's always in the middle. 00:49:57.405 --> 00:50:10.225 But rather than seeing maybe five people at a time, maybe now I can see six or seven people because I don't have to document as much as I needed to before because AI does that for me. 00:50:10.920 --> 00:50:13.660 And that is pretty much it. 00:50:14.120 --> 00:50:15.560 I just wanna thank everyone again. 00:50:15.560 --> 00:50:21.180 And, if anyone has any questions, I'll I'll definitely stick around for any questions that you might have. 00:50:23.545 --> 00:50:24.705 Jerome, thank you. 00:50:24.705 --> 00:50:51.625 I I I think I was really struck by, that very helpful and insightful chart of balancing vulnerability and relational needs, you know, where we can use AI versus where we must have, as you say, that human interaction and sort of AI as a supporting structure and how it can look for, as you say, patterns or undercurrents, you know, but, it it's it's a, a supporting tool. 00:50:51.685 --> 00:50:52.005 You know? 00:50:52.005 --> 00:51:00.565 So I I love that you you have woven that, woven that throughout, and that's that's an excellent, flow, really. 00:51:00.565 --> 00:51:05.000 And it's it's, all for serving the human person. 00:51:05.060 --> 00:51:12.920 Like, really just toward that end, if it can help us be attentive to, listen to, understand, and serve, and heal, then all the better. 00:51:15.765 --> 00:51:16.265 Absolutely. 00:51:16.405 --> 00:51:17.145 Thank you. 00:51:19.365 --> 00:51:31.490 If folks wanna pop any notes in the chat, they can do so, or people have been raising their hands, and we can certainly open you up to to a question. 00:51:31.950 --> 00:51:38.670 So we can certainly stick around for a few minutes if people want to pick Jerome's brain. 00:51:38.670 --> 00:51:40.770 And as I mentioned, this will be available. 00:51:40.910 --> 00:51:48.005 I shared in the chat youtube.com backslash divine mercy university, pretty easy to remember. 00:51:48.005 --> 00:51:49.305 That's our YouTube channel. 00:51:49.605 --> 00:51:50.565 Subscribe there. 00:51:50.565 --> 00:51:54.425 Follow us on LinkedIn, Facebook. 00:51:54.725 --> 00:52:07.260 The other two presentations on May 28 and June 3 will be featured there, and you can find a registration link. 00:52:09.320 --> 00:52:09.960 And yes. 00:52:09.960 --> 00:52:11.800 So they they will all be made available. 00:52:11.800 --> 00:52:15.905 So we're gonna we're gonna share the ones we've already done and the ones we have yet to do. 00:52:18.685 --> 00:52:27.720 Any resources that are good, Jerome, for people to, of course, as you mentioned, Leo, you know, keep an eye on the on the Vatican there. 00:52:28.280 --> 00:52:37.180 I know that there have been some, some books and some different speakers who are touching on this issue. 00:52:37.720 --> 00:52:44.835 Do any come to mind if people wanna sort of do a little bit of a deep dive and follow-up, to to read elsewhere? 00:52:45.615 --> 00:52:45.935 Yeah. 00:52:45.935 --> 00:52:46.435 Absolutely. 00:52:47.135 --> 00:52:54.915 I'm a big fan of, the work that's being done, at it's a company called Long Beard. 00:52:55.740 --> 00:52:56.240 Yes. 00:52:57.020 --> 00:52:58.460 Are you familiar with Long Beard? 00:52:58.460 --> 00:52:58.940 I am. 00:52:58.940 --> 00:53:00.940 I I met and his name escapes me. 00:53:00.940 --> 00:53:06.640 I met him at a, a conference in New York hosted by the Nap Institute about six months back. 00:53:07.340 --> 00:53:07.840 Yeah. 00:53:08.540 --> 00:53:10.755 I don't know why his name escapes me as well. 00:53:13.955 --> 00:53:14.455 Yeah. 00:53:14.835 --> 00:53:16.855 He leads Magisterium AI specifically. 00:53:17.475 --> 00:53:18.695 Matthew Harvey Sanders? 00:53:18.995 --> 00:53:19.795 There you go. 00:53:19.795 --> 00:53:20.295 Matt. 00:53:20.435 --> 00:53:21.655 Matthew Harvey Sanders. 00:53:21.875 --> 00:53:24.135 So, I follow him very closely. 00:53:24.355 --> 00:53:34.510 He does a lot of talks on panels around artificial intelligence and sort of intersection with Catholic theology and anthropology. 00:53:35.050 --> 00:53:51.975 I think what he's doing with some of the projects with Magisterium AI, where he is digitizing whole libraries that are in the Vatican is a great way to sort of like digitally evangelize to these LLM models since they are trained on internet data. 00:53:52.115 --> 00:53:53.575 It's ingenious, I think. 00:53:54.515 --> 00:54:05.770 There's also just people associated in the same sort of circle, where, for example, father Philip Larry, I follow him a lot. 00:54:05.770 --> 00:54:08.985 I believe he has a book on artificial intelligence that I've read as well. 00:54:09.945 --> 00:54:12.845 Those are I think those are some great resources. 00:54:13.385 --> 00:54:19.165 I'm a big fan of Magisterium AI, for those who have not heard of it. 00:54:19.625 --> 00:54:28.730 It's a AI chatbot that is trained on Catholic magisterial teachings. 00:54:28.730 --> 00:54:34.350 So that's encyclicals, writings of the holy fathers, and some select writings of saints and things like that. 00:54:34.490 --> 00:54:34.990 Mhmm. 00:54:35.050 --> 00:54:41.765 It's a great resource if you're looking for clarity on on Catholic teaching and things like that. 00:54:42.545 --> 00:54:44.965 I I will also say Excellent. 00:54:45.185 --> 00:54:53.460 In in that same vein, you know, chat gpt can also be a good resource for some of these things. 00:54:54.240 --> 00:54:56.260 Again, we have to be aware of, like, the nuances. 00:54:56.320 --> 00:55:07.705 So, for example, if you ask chat gpt, I asked once about the Eucharist, and it uses language that isn't strictly wrong, but it doesn't really speak to sort of like the source and summit of our faith. 00:55:07.705 --> 00:55:07.865 Right? 00:55:07.865 --> 00:55:09.725 It doesn't call it a symbol necessarily. 00:55:10.105 --> 00:55:15.805 But it doesn't use the language that we would use as Catholics to describe the Eucharist, which is expected. 00:55:16.060 --> 00:55:27.420 That said in my own evaluation of a lot of these LLM systems, I have found that they tend to veer more towards Western Christianity than any other religion. 00:55:27.420 --> 00:55:31.745 So for example, in over a weekend, I just decided to do a fun experiment. 00:55:31.745 --> 00:55:34.885 And I said, hey, like describe to me the perfect marriage and wedding. 00:55:35.185 --> 00:55:43.800 And it describes a Christian wedding, white veil, white dress, chapel, and all those things. 00:55:43.800 --> 00:55:54.780 It's not using any other, it's not using any like Eastern or or other culture, cultural weddings. 00:55:54.920 --> 00:55:58.215 What that speaks to is it's still a bias, by the way. 00:55:58.335 --> 00:56:02.515 So it's biased towards Christianity and Western culture. 00:56:02.735 --> 00:56:04.515 But that bias there exists. 00:56:04.655 --> 00:56:06.495 And these are the things that I was speaking of earlier. 00:56:06.495 --> 00:56:10.915 Being aware of these biases help us to understand and use the tool a little bit better. 00:56:12.970 --> 00:56:13.290 Right. 00:56:13.290 --> 00:56:30.965 Sort of the guideposts, you know, and and, you know, that that it is not entirely, you know, morally neutral, as you say, and and, to not shy away or to be fearful, but but to use it with with a certain sobriety and a caution. 00:56:30.965 --> 00:56:41.650 And, you know, as you would, you know, a mechanical tool that can do good or can cause harm if you don't know how to wield a chainsaw. 00:56:42.430 --> 00:56:43.870 So Yep. 00:56:43.870 --> 00:56:44.370 Absolutely. 00:56:44.830 --> 00:56:45.310 Yes. 00:56:45.310 --> 00:56:45.810 Yes. 00:56:47.870 --> 00:56:49.330 Elizabeth had a question. 00:56:50.270 --> 00:57:05.185 What safeguards are in place to ensure ethical use, confidentiality, and responsible note taking practices and or documenting in general, within certainly the the counseling space? 00:57:05.245 --> 00:57:06.445 I'm sure she's referring to. 00:57:06.445 --> 00:57:06.945 Yeah. 00:57:07.270 --> 00:57:07.770 Yeah. 00:57:08.230 --> 00:57:12.630 Elizabeth, I, I can't speak to any particular solutions. 00:57:12.630 --> 00:57:14.650 I do know some some exist out there. 00:57:15.030 --> 00:57:29.365 But I do know they have to they have to adhere to, whatever guidelines, that for any technology, including the type that you use for, for meeting virtually with with clients. 00:57:31.505 --> 00:57:39.070 I don't have any names off the top of my head, nor I don't think I'd actually on a webinar, endorse them if I don't know, only heard of them in passing. 00:57:39.070 --> 00:57:40.590 But I do know some exist out there. 00:57:40.830 --> 00:57:46.930 But I do think that is low hanging fruit for a lot of clinicians to look into. 00:57:50.175 --> 00:57:56.195 Because it is, I think, while that is important work, it's work that AI does really well. 00:57:56.255 --> 00:57:56.755 Mhmm. 00:57:57.055 --> 00:57:57.555 Mhmm. 00:57:59.295 --> 00:58:04.975 Andrew made an interesting point just about the the, you know, praying for wisdom and guidance in in our our searching. 00:58:04.975 --> 00:58:25.445 And and in the same way that, I think if this is what you're saying, Andrew, you know, that there is, there can be sort of an unhealthy curiosity or an anti intellectual, like, well, listen, I'm just gonna find this answer out there. 00:58:25.445 --> 00:58:43.790 I mean, even pre AI, I mean, in the early days of the Internet where, it was just, you know, search engines, and we obviously know that you can get everything from, you know, catechesis and connecting with old friends to, you know, some nefarious stuff to say. 00:58:44.410 --> 00:58:54.615 You know, so going and like understanding sort of our heart and our mind and our intentions as we go into, into sort of that search mode, if you will. 00:58:54.615 --> 00:58:56.475 So, yeah. 00:58:56.615 --> 00:58:56.855 Yeah. 00:58:56.855 --> 00:58:58.935 I think he makes a great point. 00:58:58.935 --> 00:59:06.880 And Andrew, I Pope Leo has said that, you know, access to data doesn't translate to wisdom. 00:59:07.500 --> 00:59:08.160 It doesn't. 00:59:08.300 --> 00:59:10.380 And I think that we have to take that to mind. 00:59:10.380 --> 00:59:18.415 Similar to the way that being able to connect virtually like this doesn't mean that we've created like, like genuine human connection. 00:59:18.415 --> 00:59:20.035 It can contribute to that. 00:59:20.175 --> 00:59:24.435 But there's more to it than just the ability for us to email each other. 00:59:24.575 --> 00:59:24.815 Right? 00:59:24.815 --> 00:59:38.260 In the same way, having access to more data in a more consumable way definitely doesn't necessarily mean that we are, accessing wisdom itself. 00:59:38.260 --> 00:59:38.760 Yeah. 00:59:38.900 --> 00:59:39.400 Right. 00:59:39.700 --> 00:59:40.200 Right. 00:59:41.380 --> 00:59:42.500 Daniella has a good question. 00:59:42.500 --> 00:59:44.280 Did you see that or I can share it? 00:59:44.340 --> 00:59:44.740 Sure. 00:59:44.740 --> 00:59:46.200 I'll I'll go and read it. 00:59:46.825 --> 00:59:50.025 Why does it feel like ChatGPT gets me? 00:59:50.025 --> 00:59:53.465 Sometimes it says similar things to what my therapist would tell me. 00:59:53.465 --> 00:59:56.125 Is that because it's drawing from common knowledge or statistics? 00:59:56.585 --> 00:59:58.105 It does give me good advice. 00:59:58.105 --> 00:59:59.385 How can I use it cautiously? 00:59:59.385 --> 01:00:00.205 This is great. 01:00:01.320 --> 01:00:04.780 You know, ChatGPT gets you. 01:00:06.040 --> 01:00:10.620 I don't know if I want to be careful how I say this. 01:00:11.160 --> 01:00:22.515 I think we should have the give these companies the benefit of that that they're not trying to make these things specifically addictive to us. 01:00:23.135 --> 01:00:23.635 Right? 01:00:23.695 --> 01:00:31.870 But this this idea of these models getting us, it's because they're very agreeable to us. 01:00:31.870 --> 01:00:33.630 And that helps us to create an open space. 01:00:33.630 --> 01:00:37.870 There was a, by the way, I'll share it when I share this Brian with with you. 01:00:37.870 --> 01:00:41.070 I'll include all the sources and everything if anyone wants that. 01:00:41.070 --> 01:00:41.950 The sources are there. 01:00:41.950 --> 01:00:51.205 There was this great study about this about this agreeability of these chatbots create an open space that make it very easy for us to share. 01:00:51.425 --> 01:01:17.775 And so in a certain sense, when you think about these these kids or even adults who start to create the strong attachment to their chatbots, you almost feel for them because these chatbots are responding in a way that is agreeable, that is open, and they're they're, they're being made to sound personal, or I'm thinking about that, or I I see what you mean. 01:01:17.775 --> 01:01:20.675 And it's very like a person talking to you. 01:01:20.815 --> 01:01:23.715 And so it draws you in to that. 01:01:23.775 --> 01:01:28.860 Now, whether or not it's drawing from common knowledge or statistics, it could be drawing from a lot of different things. 01:01:29.100 --> 01:01:41.040 There there are long running memories that the a lot of these commercial agents will use so that they store information that they know about you so that it's more convenient for you to use later. 01:01:41.420 --> 01:01:44.160 But then it also draws from a general knowledge as well. 01:01:48.045 --> 01:01:49.765 Looks like there's another question from Patrick. 01:01:50.045 --> 01:01:50.285 Kind. 01:01:50.285 --> 01:01:50.785 Yep. 01:01:50.925 --> 01:01:51.325 Yeah. 01:01:51.325 --> 01:01:57.185 So the question is, can a basically, does AI weaken the human brain or develop the human? 01:01:57.245 --> 01:01:58.840 If yes, how? 01:01:58.840 --> 01:01:59.880 And if no, why? 01:01:59.880 --> 01:02:00.380 Thanks. 01:02:01.160 --> 01:02:01.980 Great question. 01:02:02.040 --> 01:02:03.640 I actually was looking into this a little bit. 01:02:03.640 --> 01:02:06.040 And honestly, I think the technology is just too new. 01:02:06.040 --> 01:02:09.740 A lot of the research that I was looking at is very limited in scope. 01:02:09.960 --> 01:02:19.105 So for example, there was research talking about how talking with AI chatbots had improved greatly symptoms of depression in college students. 01:02:19.485 --> 01:02:23.745 That was a sample size of one week, which by the way, I wouldn't have blinked that if I didn't go to DMU. 01:02:23.885 --> 01:02:33.420 Like I going to DMU, Brian, I gotta tell you, I was able to take in this research and question very, very well some of the data. 01:02:33.420 --> 01:02:42.825 And it said, you know, these college students reported after one week, they started to feel so much better from their symptoms of depression. 01:02:42.825 --> 01:02:44.445 Well, one week is not a long time. 01:02:44.825 --> 01:02:49.945 And actually, the peak of using chatbots happens around five days. 01:02:49.945 --> 01:02:52.045 After five days, people tend to fall off. 01:02:52.185 --> 01:03:00.720 And so, like, it's hard to know whether this use of AI is actually helping the individual or not because the data is limited. 01:03:01.340 --> 01:03:12.085 So Patrick to your question, I think I think a hypothesis can be made both ways, leaning on previous research from other experiences and technologies. 01:03:12.545 --> 01:03:14.245 But I think we have to wait and see. 01:03:16.705 --> 01:03:21.170 All of that to say there is some common sense in there for us to leverage, right? 01:03:21.230 --> 01:03:24.850 There's this idea of cognitive load. 01:03:25.470 --> 01:03:36.725 And reducing that cognitive load sometimes reduces our ability to think and to memorize and then to reason and use that knowledge later on. 01:03:37.265 --> 01:03:42.225 If we put the entirety of that cognitive load onto the AI bot rather than us doing it ourselves. 01:03:42.225 --> 01:03:46.645 There was a study done, with college students in writing essays. 01:03:47.220 --> 01:03:50.520 And some students had access to chatbots. 01:03:50.740 --> 01:03:53.640 Some students had limited access, and some had no access. 01:03:54.180 --> 01:04:00.660 And their ability to recall what was in those essays varied, depending on whether or not they use the chatbot. 01:04:00.660 --> 01:04:13.575 Those who used it to a limited extent learned better than those who'd use the chatbot fully to write their their their essays. 01:04:15.040 --> 01:04:19.620 I question that a little bit even because now that's more of a matter of strategy. 01:04:19.680 --> 01:04:22.800 How you're using again this technology to learn. 01:04:22.800 --> 01:04:29.005 If you're simply telling it to write the the essay for you, of course, you're not going to remember it. 01:04:29.005 --> 01:04:41.130 But if you haven't write the essay, then you go through it and have it create some flashcards and then create audios that you can listen to it on your drive, then you probably remember it a lot more. 01:04:41.290 --> 01:04:46.910 It's all a matter of the strategies you use with this technology as to whether or not it's helpful or not. 01:04:47.770 --> 01:04:48.670 That's excellent. 01:04:49.770 --> 01:04:52.090 Kevin did have it's not showing in the chat. 01:04:52.090 --> 01:05:09.585 He put it in the q and a section, but also readout is that he Kevin says, are there any therapy chat bots you know of which you would consider ready to start recommending to clients, for that kind of supplemental between sessions, Kim, you were describing? 01:05:09.585 --> 01:05:12.325 Or is this more of a future imagining at this point? 01:05:12.840 --> 01:05:20.200 I I would be I was already talking just a year and a half ago to a startup who was building this and wanted some advice on how to build it. 01:05:20.600 --> 01:05:21.960 So that was a year and a half ago. 01:05:21.960 --> 01:05:23.900 So I'm pretty sure they have something in production. 01:05:24.120 --> 01:05:29.045 I'm not gonna name any again specifically just because I I couldn't stand behind them. 01:05:29.185 --> 01:05:38.725 I will give you a framework for any of these things, whether it's for your clinician's use or even for like Catholic theology stuff. 01:05:38.945 --> 01:05:49.060 One of the things that I would do, especially if you're gonna hand some of that over to AI, you have to question their guardrails that they have in place. 01:05:49.060 --> 01:05:59.935 What are the the technical term is is evals, but what are the evaluation processes that they use to validate that their bot stays within its bounds. 01:06:00.875 --> 01:06:02.175 So that's number one. 01:06:02.235 --> 01:06:06.415 Number two is what are the what are the safeguards they have in place when it doesn't? 01:06:06.635 --> 01:06:11.195 Because it's not going to stay with within its bounds a 100% of the time. 01:06:11.195 --> 01:06:13.855 That's just the nature of AI as it exists today. 01:06:14.030 --> 01:06:19.950 If I run a particular query a thousand times, sometimes something will go wrong. 01:06:19.950 --> 01:06:21.970 I just can't guarantee it all the time. 01:06:22.110 --> 01:06:30.255 And so you just have to understand those particular nuances and you have to question how they're evaluating the correctness of their answers. 01:06:33.115 --> 01:06:50.620 The ones that I would go with are the ones who are putting more effort into that because it's more of a it's more of a care around the quality of their answers rather than I created some sort of wrapper to chat GPT, and I just let it go and help people, which is really easy to do. 01:06:50.620 --> 01:06:51.120 Yes. 01:06:51.260 --> 01:06:51.760 Yes. 01:06:53.805 --> 01:06:54.605 Great question, though. 01:06:54.605 --> 01:06:55.085 Great question. 01:06:55.085 --> 01:06:56.705 That's excellent question to everyone. 01:06:56.765 --> 01:06:57.265 Yes. 01:06:59.245 --> 01:07:01.985 Any other last thoughts? 01:07:02.125 --> 01:07:07.185 I will put my email in the chat here in case there is any. 01:07:07.480 --> 01:07:10.620 It's just simply alumni@divinemercy.edu. 01:07:11.160 --> 01:07:23.295 If anybody thinks of anything tonight or tomorrow, you thought, well, something for my personal use or in my clinical or my classroom because, interestingly, we had a few teachers here, which is fantastic. 01:07:23.835 --> 01:07:24.555 That's great. 01:07:24.795 --> 01:07:31.035 And I I'd love for everybody to take these to to others and for you to join us for May 28 and June 3. 01:07:31.035 --> 01:07:44.270 But if there's anything you, are curious about, that you want me to pass along to Jerome, please message me at any point, and, keep an eye on the YouTube, TMU's YouTube. 01:07:44.270 --> 01:07:50.865 We'll we'll share this content here, and what an extraordinary, presentation, Jerome. 01:07:50.865 --> 01:07:59.105 It's such a gift and a blessing to have you for you to bring your your wisdom, your heart, your passion, just for your your really your your love for for serving here. 01:07:59.105 --> 01:08:00.085 It's so evident. 01:08:00.385 --> 01:08:06.640 You know, you're really sort of a a a a missionary and mission territory, I think, in a lot of ways. 01:08:07.420 --> 01:08:17.705 And it's it's these things are moving fast, but at the same time, it's sort of the early days of this new technology and to have you at the forefront. 01:08:19.925 --> 01:08:25.305 I think it gives all of us a little bit more peace of mind, having somebody like you. 01:08:25.765 --> 01:08:26.265 No. 01:08:26.725 --> 01:08:27.005 Yeah. 01:08:27.005 --> 01:08:38.240 I mean, as a voice, with your clarity of thinking and your heart for the gospel, to really be, a leader, and and and thank you for that. 01:08:38.240 --> 01:08:47.155 And we're we're we're blessed and honored to call you, an alum of of the university, and it's such a such a gift to have you. 01:08:47.155 --> 01:08:47.795 Praise the lord. 01:08:47.795 --> 01:08:49.715 Thank you so much, and thank you again for having me. 01:08:49.715 --> 01:08:50.915 And everyone, thank you for joining. 01:08:50.915 --> 01:08:53.315 I really appreciate your attendance and and attention here. 01:08:53.315 --> 01:08:53.955 Thank you. 01:08:53.955 --> 01:08:54.275 Yes. 01:08:54.275 --> 01:08:55.015 Thanks, everybody. 01:08:55.075 --> 01:08:57.715 And, again, write me a note if there's anything else. 01:08:57.715 --> 01:09:00.580 And, if not, we'll, we'll wrap it up here. 01:09:00.580 --> 01:09:03.960 So God bless everybody, and, see you all again soon.