WEBVTT - Detecting Depression with AI 0:00:08.000 --> 0:00:10.320 I know I need to get ready for work, but 0:00:10.400 --> 0:00:14.280 I'm just so so tired. Maybe I can skip work 0:00:14.400 --> 0:00:19.880 to take a nap justice once. What's that? Oh, my 0:00:19.960 --> 0:00:23.520 depression monitors detecting early signs of a depressive state. I 0:00:23.560 --> 0:00:26.680 thought it was just tired, but this might be a 0:00:26.720 --> 0:00:30.280 bigger issue. Let me set in an appointment with my therapist. 0:00:40.960 --> 0:00:44.120 Hey there, I'm grain class and this is technically speaking 0:00:44.360 --> 0:00:47.640 an Intel podcast. The show is dedicated to highlighting the 0:00:47.680 --> 0:00:51.560 way technology is revolutionizing the way we live, work and move. 0:00:52.479 --> 0:00:55.080 In every episode, we'll connect with innovators in areas like 0:00:55.200 --> 0:00:59.440 artificial intelligence to better understand the human centered technology they've developed. 0:01:00.600 --> 0:01:04.160 Mental health care solutions remain underinvested in many communities around 0:01:04.160 --> 0:01:06.960 the world, yet so many suffer from issues that they 0:01:06.959 --> 0:01:10.200 don't even know. They have a lot of these conversations 0:01:10.240 --> 0:01:13.120 around health care, hinge on making therapy more accessible to 0:01:13.160 --> 0:01:17.400 those in need. However, it can be difficult to determine 0:01:17.440 --> 0:01:22.120 that one is experiencing depression or mental health crisis. Artificial 0:01:22.160 --> 0:01:25.440 intelligence is at the forefront of many different advancements in healthcare, 0:01:26.200 --> 0:01:28.400 but today we are going to dive into how it 0:01:28.520 --> 0:01:32.280 is working to make mental health care more accessible. To everyone. 0:01:33.080 --> 0:01:35.240 In order to do that, I have to introduce you 0:01:35.400 --> 0:01:41.120 to a special guest joining me now is Tinasawho. Tina 0:01:41.120 --> 0:01:44.000 Saho was a high school student when she started exploring 0:01:44.040 --> 0:01:47.880 coding and engineering with AI. She never considered herself to 0:01:47.920 --> 0:01:51.600 be overly interested in technology to start. However, she took 0:01:51.600 --> 0:01:54.480 her natural curiosity and eventually was invited to be a 0:01:54.480 --> 0:01:57.800 part of Intel's AI for Youth Pilot program, where she 0:01:57.840 --> 0:02:01.200 developed a tool that uses AI to detect and predict 0:02:01.240 --> 0:02:05.160 patterns of depression with around eighty percent accuracy. Since then, 0:02:05.320 --> 0:02:08.240 she's been awarded the Calm Trust Fellowship for Women. She 0:02:08.280 --> 0:02:11.720 currently attends to Ayabata College at the University of Delhi, 0:02:12.160 --> 0:02:15.679 where she's completing a bachelor's in computer science and exploring 0:02:15.720 --> 0:02:18.680 more opportunities to use AI as a tool for mental 0:02:18.680 --> 0:02:19.480 health and more. 0:02:19.960 --> 0:02:22.840 Welcome to the show, Tina, Hi, thank you for inviting 0:02:22.880 --> 0:02:25.679 me to this show, and I'm feeling more than privileged 0:02:25.720 --> 0:02:28.280 to be a part of this podcast. So numbers stay 0:02:28.320 --> 0:02:28.880 to everyone. 0:02:33.360 --> 0:02:36.040 I'm really interested in your story of how you got 0:02:36.080 --> 0:02:40.200 into the STEM field, what prompted your interest in that 0:02:40.280 --> 0:02:41.760 field and also in AI. 0:02:42.360 --> 0:02:44.720 To begin with, till my tenth class, I was a 0:02:44.760 --> 0:02:47.920 student or a person who was always against technology, like 0:02:48.120 --> 0:02:50.519 I always found in fact that the negative impacts that 0:02:50.600 --> 0:02:53.880 technology brings and whatever ethical concerns are there, they cannot 0:02:53.880 --> 0:02:56.360 be solved and they are simply too much to be 0:02:56.400 --> 0:03:00.000 on the side of technology. But I remember in twenty nineteen, 0:03:00.400 --> 0:03:03.080 the AAFO Youth program was launched and it was launched 0:03:03.120 --> 0:03:06.160 in our school, which is Salvangal Senior Secondary School. So 0:03:06.600 --> 0:03:09.919 when I participated in that program, I got to know 0:03:09.960 --> 0:03:14.720 what artificial intelligence is, how many astonishing and stounding possibilities 0:03:14.760 --> 0:03:18.520 AI can unlock and has already unlocked. And therefore, this 0:03:18.639 --> 0:03:22.600 program pivoted my you know journey, my career, my professional journey, 0:03:22.800 --> 0:03:24.919 and my cademic journey from being a non tech student 0:03:24.960 --> 0:03:28.160 to becoming a tech student finally pursuing computer science as 0:03:28.320 --> 0:03:29.040 my graduation. 0:03:29.960 --> 0:03:35.200 Okay, was there a particular topic or person that really 0:03:35.240 --> 0:03:38.080 did spark that interest? What was the particular topic that 0:03:38.160 --> 0:03:40.240 actually got you really fired up? 0:03:40.880 --> 0:03:43.200 So in my school there were two teachers who actually 0:03:43.280 --> 0:03:46.760 you know, molded my thinking and they actually infused this 0:03:46.840 --> 0:03:49.640 critical thinking aspect in me and they opened me to 0:03:49.680 --> 0:03:53.840 the world of possibilities that science, that technology and engineering offers. 0:03:54.080 --> 0:03:55.800 And I would like to take their name as well 0:03:55.800 --> 0:03:58.040 as a you know, token of respect. So it's one 0:03:58.120 --> 0:04:02.080 that Mam and Serbiam they empowered me. They encouraged me 0:04:02.160 --> 0:04:04.360 to think beyond what I see. 0:04:04.920 --> 0:04:09.160 And that led you to the Intel's AI for Youth program. 0:04:09.760 --> 0:04:12.640 What sort of projects have you worked on? Are you 0:04:12.760 --> 0:04:13.800 working on right now? 0:04:14.480 --> 0:04:17.760 This isn't my participating in this program. I build Happiness Guru, 0:04:17.800 --> 0:04:21.120 which is a model that predicts depression. Apart from that too, 0:04:21.240 --> 0:04:24.440 I was also a part of Utul Tinkling Labs. Through 0:04:24.480 --> 0:04:27.560 the Detail Lab of our school. I basically got to 0:04:27.600 --> 0:04:31.239 know about Intel's AfOR Youth program only and in there itself, 0:04:31.320 --> 0:04:33.320 I build a few projects and one of them was 0:04:33.360 --> 0:04:34.119 Happiness Guru. 0:04:34.720 --> 0:04:38.720 And with the Happiness Crew, is that related to the 0:04:38.760 --> 0:04:44.839 depression detection research you've been doing? What kind of prompted 0:04:44.880 --> 0:04:47.200 you to look in that direction in the field of 0:04:47.880 --> 0:04:51.840 depression and then using technology? Because generally, speaker, we don't 0:04:51.880 --> 0:04:57.320 associate technology with treating depression or even detecting depression. So 0:04:57.400 --> 0:04:59.280 what was the spark for you there? 0:05:00.240 --> 0:05:02.440 So while I was in this program, I was in 0:05:02.480 --> 0:05:05.600 Class ten and my Plus ten results were out and 0:05:05.680 --> 0:05:08.240 they were not as much as I expected, and I 0:05:08.400 --> 0:05:12.120 went into a phase of depression because I associated myself 0:05:12.160 --> 0:05:15.640 worth with the Marxi score, so my own personal experience 0:05:15.680 --> 0:05:19.320 of dealing with depression. And then at that time, you 0:05:19.360 --> 0:05:21.919 know that society rates were very alarming among the youth, 0:05:22.040 --> 0:05:25.440 especially aged between fifteen to twenty nine. And we found 0:05:25.480 --> 0:05:28.600 that the driving forces behind you know these when these 0:05:28.640 --> 0:05:32.799 societies were you know, peer pressure, overburdening academics, financial stress, 0:05:33.320 --> 0:05:36.039 and too much expectations that we have from youth, you know, 0:05:36.120 --> 0:05:38.839 especially if you talk about teenage and someone who is 0:05:39.279 --> 0:05:42.320 in between eighteen to twenty five. So on researching, we 0:05:42.360 --> 0:05:45.440 found that these societ rates are very alarming, they're very distressing, 0:05:45.480 --> 0:05:48.240 and these are the leading cause of the depression or stress. 0:05:48.600 --> 0:05:51.159 And thereby we thought that we must come up with 0:05:51.200 --> 0:05:54.880 some solution that can basically help us predict which person 0:05:54.920 --> 0:05:56.880 is going through depression, and that to in a very 0:05:56.920 --> 0:06:00.640 human friendly manner, not making someone uncomfortable with the kind 0:06:00.680 --> 0:06:02.560 of procedures or with the kind of system we have. 0:06:03.160 --> 0:06:07.000 So these were the I would say, the enablers that 0:06:07.120 --> 0:06:09.320 led our team building this solution. 0:06:09.920 --> 0:06:13.440 And in terms of the happiness grew app can you 0:06:13.520 --> 0:06:16.120 just explain how it actually works, you know, to try 0:06:16.120 --> 0:06:18.480 and detect the early signs of depression. 0:06:19.279 --> 0:06:23.160 First of all, it's our web based application. While building 0:06:23.200 --> 0:06:26.080 this project, the queue that we took, you know, to 0:06:26.160 --> 0:06:29.080 build the entire model was that after our research, we 0:06:29.160 --> 0:06:32.160 got to know that a person's vocabulary can be a 0:06:32.200 --> 0:06:35.800 mirror into their mental state. And taking this as the queue, 0:06:36.080 --> 0:06:39.640 we build this project which tries to analyze the emotional 0:06:39.760 --> 0:06:42.960 quotient of a person of a user through their facial 0:06:43.000 --> 0:06:46.640 expression and then their textual responses that the user is 0:06:46.640 --> 0:06:49.120 going to provide to the AA machine. So the working 0:06:49.160 --> 0:06:51.680 of the project is divided into three steps. The first 0:06:51.720 --> 0:06:54.919 step is emotion detection stage, and in this stage you 0:06:55.040 --> 0:06:57.760 basically need to stand in front of your laptop or 0:06:57.800 --> 0:07:00.720 whatever device you are using this web application, and then 0:07:00.760 --> 0:07:03.200 it detects your current mode, whether you're happy or sad, 0:07:03.200 --> 0:07:06.279 you're neutral, angry. Then the next step is that user 0:07:06.360 --> 0:07:10.280 is asked to answer nine questions and there's a scale 0:07:10.280 --> 0:07:13.280 of relevance and then they need to select how much 0:07:13.320 --> 0:07:15.560 relevant or how much they are able to relate this 0:07:15.800 --> 0:07:19.000 to this situation. Then, after these two steps, a threshold 0:07:19.080 --> 0:07:21.720 score is generated which gives the initial lead. If the 0:07:21.760 --> 0:07:23.960 person is stressed or not, and if the score is 0:07:23.960 --> 0:07:26.240 below the threshold that we have said, the person is 0:07:26.280 --> 0:07:29.320 predicted as happy, while in the other case, user is 0:07:29.360 --> 0:07:31.200 taken to the third step, which is the final step. 0:07:31.400 --> 0:07:34.160 And this step consists of four descriptive questions which he 0:07:34.320 --> 0:07:36.960 or she can use as a platform to went out 0:07:37.080 --> 0:07:40.600 all his or her feelings and thoughts. So whatever answers 0:07:40.720 --> 0:07:44.200 user will give to these four descriptive questions, these answers 0:07:44.200 --> 0:07:47.200 will be used as a basis of classification. Then the 0:07:47.240 --> 0:07:49.560 machine will predict whether the user is depressed or not. 0:07:49.680 --> 0:07:52.480 So this AI machine, whatever you know input we are 0:07:52.480 --> 0:07:55.360 giving in this step. There's a model namely SVM, which 0:07:55.360 --> 0:07:58.760 is support vector machine. It's a non contextual classification model. 0:07:58.760 --> 0:08:02.120 It is basically used to classify things. And then we 0:08:02.240 --> 0:08:04.360 are using this model on the kind of you know, 0:08:04.480 --> 0:08:07.040 language or keywords that are used in the answers. And 0:08:07.080 --> 0:08:09.760 then accordingly the results are given out that whether the 0:08:09.760 --> 0:08:12.000 person is stressed or not, and if the person is stressed, 0:08:12.480 --> 0:08:16.520 automatically the person is consulted to the concerned authorities or 0:08:16.560 --> 0:08:20.600 counselor otherwise the person is predicted as happy or not stressed. 0:08:22.440 --> 0:08:26.320 Detecting and treating mental health is something with which many 0:08:26.400 --> 0:08:30.640 societies around the world struggle. According to the World Health Organization, 0:08:31.160 --> 0:08:33.880 approximately two hundred and eighty million people in the world 0:08:33.920 --> 0:08:37.160 suffer from depression and more than three hundred million are 0:08:37.200 --> 0:08:41.479 living with anxiety. Many people with these mental health conditions 0:08:41.559 --> 0:08:46.000 exhibit some symptoms as children or young adults, but based 0:08:46.040 --> 0:08:49.000 on guidance from the US National Institute of Mental Health, 0:08:49.800 --> 0:08:53.560 depression can only be diagnosed once an individual exhibits the 0:08:53.640 --> 0:08:57.400 five major symptoms of depression every day, all day for 0:08:57.440 --> 0:09:00.680 a minimum of two weeks. Imagine how we could help 0:09:00.720 --> 0:09:04.280 people earlier if we were able to identify depression with 0:09:04.400 --> 0:09:07.200 the help of AI tools like the Happiness Guru model. 0:09:10.120 --> 0:09:14.600 How does one actually create that model? What data is 0:09:14.760 --> 0:09:18.760 needed to train that model so that it can get 0:09:18.760 --> 0:09:19.440 that output. 0:09:20.040 --> 0:09:23.120 So basically, whenever we build any project, we were taught 0:09:23.120 --> 0:09:25.120 this thing in the program itself that there's a whole 0:09:25.120 --> 0:09:27.840 project cycle that needs to be taken into you know, 0:09:27.840 --> 0:09:30.959 account while we're building any project. So the first step 0:09:31.040 --> 0:09:33.800 that comes into the AA project cycle is problem scoping. 0:09:33.840 --> 0:09:36.560 So we have problem statements, we have a stakeholders, and 0:09:36.760 --> 0:09:39.040 we have our ideal solution as well. Now comes data 0:09:39.080 --> 0:09:42.600 acquisition so basically to make this project work the way 0:09:42.640 --> 0:09:45.280 it is working right now, data was collected you know, 0:09:45.320 --> 0:09:48.679 anonymously through offline and online surveys and across five different 0:09:48.679 --> 0:09:52.680 schools across India. So during these surveys, we briefed students 0:09:52.720 --> 0:09:55.320 in the school what this survey is about and then 0:09:55.360 --> 0:09:58.040 they were asked to fill out that form which contained 0:09:58.120 --> 0:10:01.400 descriptive questions. Now, these descriptive questions that we selected, these 0:10:01.440 --> 0:10:04.679 were validated by a team of psychiatrists and counselors and 0:10:04.720 --> 0:10:07.160 then with the help of this survey process, we were 0:10:07.200 --> 0:10:09.640 able to develop an authentic data set of seven hundred 0:10:09.679 --> 0:10:13.880 plus centuries where the students basically wrote whatever they felt 0:10:13.920 --> 0:10:16.400 during that time and you know, went out their thoughts 0:10:16.480 --> 0:10:19.360 in that survey. The responses were labeled the on the 0:10:19.360 --> 0:10:21.880 scale of A two D, with A being least sever 0:10:22.120 --> 0:10:26.240 like perfectly healthy mentally and to D being needing immediate 0:10:26.280 --> 0:10:29.040 support from professionals and family. And this was done with 0:10:29.080 --> 0:10:32.000 the help of our school counselor, Ishitan Atara, So she 0:10:32.080 --> 0:10:35.400 helped us in you know, laboring these responses and then 0:10:35.480 --> 0:10:38.880 this data was used to train that SVM model that 0:10:38.960 --> 0:10:41.280 I was talking about that is a part of step three, 0:10:41.679 --> 0:10:44.719 So we need to convert this offline data into a 0:10:44.760 --> 0:10:48.439 digitized format because that's how model gets trained. So we 0:10:48.480 --> 0:10:51.280 did that, we started classifying, and then we trained the 0:10:51.400 --> 0:10:54.280 SVM model. Apart from that, there's one more thing that 0:10:54.320 --> 0:10:56.679 has went into this. The step one which I talked 0:10:56.679 --> 0:11:00.560 about is about, you know, recognizing whatever current emotion the 0:11:00.640 --> 0:11:03.760 user has, whatever their emotion is currently while they're using 0:11:03.840 --> 0:11:07.679 So this is turned with the help of library basically 0:11:07.720 --> 0:11:10.480 fast a dot Vision. So fast a dot Vision is 0:11:10.480 --> 0:11:13.280 a library that is used for computer vision tasks. And 0:11:13.320 --> 0:11:15.720 then we have trained this module using a data set. 0:11:15.880 --> 0:11:18.520 So this data set consisted of two thousand rows i 0:11:18.520 --> 0:11:22.120 would say, which consisted of facial expressions of different people, 0:11:22.240 --> 0:11:25.680 like there were videos and images of people from different 0:11:26.200 --> 0:11:30.040 genders and heritages of different backgrounds, and then they were 0:11:30.120 --> 0:11:33.360 classified as happy saturn nedle to train our module, which 0:11:33.440 --> 0:11:34.680 was fast a Dot Vision. 0:11:36.840 --> 0:11:40.120 What Tina is describing in her design philosophy is very 0:11:40.200 --> 0:11:44.000 interesting because in a way it mirrors processes used by 0:11:44.080 --> 0:11:49.720 psychiatrists and counselors to identify depression in young people at schools. However, 0:11:50.000 --> 0:11:55.319 in her system the effectiveness is amplified. Oftentimes people experiencing 0:11:55.360 --> 0:11:58.840 depression are not able to recognize the symptoms in themselves, 0:11:59.120 --> 0:12:02.520 and for young people particular, having access to a professional 0:12:02.520 --> 0:12:06.080 who could observe and identify the science is not guaranteed. 0:12:06.720 --> 0:12:11.080 For cultural, social and economic reasons, mental health is largely ignored. 0:12:11.720 --> 0:12:14.280 I can see the benefit of an automated system being 0:12:14.360 --> 0:12:17.080 used to identify it and how that can help those 0:12:17.240 --> 0:12:21.000 with reservations around mental healthcare take that crucial first step. 0:12:22.880 --> 0:12:26.440 You're listening to technically Speaking an Intel podcast will be 0:12:26.520 --> 0:12:36.679 right back. Welcome back to technically Speaking an Intel podcast. 0:12:37.000 --> 0:12:42.160 I'm here now with Tina. So, So, in terms of 0:12:42.200 --> 0:12:46.920 your research or next phase, do you think these sorts 0:12:47.320 --> 0:12:52.720 of wearable devices or things that can detect people's emotions, 0:12:53.200 --> 0:12:55.360 do you see a future where that could be a 0:12:55.400 --> 0:12:58.920 possibility where we could get in early in terms of 0:12:59.280 --> 0:13:00.720 detecting depression. 0:13:01.480 --> 0:13:03.600 Yes, there can be. In fact, there's been a rise 0:13:03.640 --> 0:13:06.160 in it lately. Like I've been following up the news 0:13:06.200 --> 0:13:09.160 around this, and I've got to know that there was 0:13:09.200 --> 0:13:12.520 some institute in New York itself which conducted a study 0:13:12.520 --> 0:13:15.560 which basically built an machine learning model that took the 0:13:15.679 --> 0:13:18.840 data of thousands users, and then this model was able 0:13:18.880 --> 0:13:21.760 to tell whether a person was mentally healthy or not. 0:13:22.280 --> 0:13:25.080 So we need to understand how this works for us 0:13:25.120 --> 0:13:28.160 to fall like, we are basically collecting data points in 0:13:28.280 --> 0:13:30.800 terms of different variables, and these variables are like you know, 0:13:31.080 --> 0:13:33.200 what at our pulse rate, what is our heart beat? 0:13:33.679 --> 0:13:36.200 And I mean different things that can be measured by 0:13:36.240 --> 0:13:39.480 these devices, by these variables to find the relation between 0:13:39.640 --> 0:13:42.600 someone's mental health and whatever data points we are collecting. 0:13:43.040 --> 0:13:46.280 So there's a possibility that in the coming year we 0:13:46.480 --> 0:13:49.040 can lead mental health care services. Apart from this, a 0:13:49.080 --> 0:13:50.920 similar thing that strucks to me right now is that 0:13:51.120 --> 0:13:55.359 brain computer interface. I mean well, brain computer interface is 0:13:55.400 --> 0:13:58.680 a machine that actually helps us to control a device 0:13:58.720 --> 0:14:02.480 or machine using our brain. So if something of that 0:14:02.559 --> 0:14:05.640 sort can be infused with machine learning, and then if 0:14:05.640 --> 0:14:08.800 we can build some solution that is oriented towards solving 0:14:08.880 --> 0:14:11.720 mental health problems that exist, that is oriented towards providing 0:14:11.720 --> 0:14:15.120 more healthcare services, like those that accessible enough and affordable 0:14:15.440 --> 0:14:18.280 as well, So I think majority of problems can be 0:14:18.360 --> 0:14:21.680 solved in this area. 0:14:21.880 --> 0:14:26.600 Tina mentioning BCI or brain computer interface reminds me of 0:14:26.600 --> 0:14:30.840 the conversation in episode three with Jaggedish and Lama. We 0:14:30.920 --> 0:14:34.720 tend to think of BCI as human brains controlling the 0:14:34.760 --> 0:14:37.920 function of a machine, like moving a mouse cursor or 0:14:37.960 --> 0:14:42.000 controlling a robotic limb. However, Tina imagines a world where 0:14:42.040 --> 0:14:45.280 our brains can simply inform machines on how to service us. 0:14:46.080 --> 0:14:47.880 It is not so much that you would need to 0:14:48.040 --> 0:14:51.200 even think about being helped, but the machine learning process 0:14:51.200 --> 0:14:53.560 would allow a tool to remind you of a service 0:14:53.600 --> 0:14:56.920 you need. It's almost like having a second brain. I 0:14:56.960 --> 0:14:59.640 can't wait to see all of the medical applications this 0:14:59.720 --> 0:15:05.160 open up to the world. Particularly through the pandemic and 0:15:05.200 --> 0:15:09.160 post pandemic, there was a rise in mental health issues 0:15:09.200 --> 0:15:14.080 which needed expert care. Now do you think that AI 0:15:14.400 --> 0:15:18.680 can play a role in actually providing therapy for people 0:15:18.720 --> 0:15:23.040 with mental health concerns. I recently read an article in 0:15:23.160 --> 0:15:27.920 Time magazine about robot, which is a AI personal therapist. 0:15:28.680 --> 0:15:31.720 I'd like to get your thoughts as to whether they 0:15:31.760 --> 0:15:36.240 could actually provide useful advice for people to help manage 0:15:36.320 --> 0:15:38.440 their depression and mental health issues. 0:15:39.400 --> 0:15:41.680 So, when we think of mental health care, you know, 0:15:41.960 --> 0:15:45.960 the corner store of this is communication. It's not depending 0:15:46.040 --> 0:15:49.400 on the procedures, but more on the communication. Like if 0:15:49.480 --> 0:15:52.000 we know that therapist and the patient that there should 0:15:52.000 --> 0:15:54.840 be a strong relationship between them. The relationship should be 0:15:54.840 --> 0:15:57.520 good enough so that the patient can communicate with their 0:15:57.560 --> 0:16:01.160 therapists and then the problem can whatever problem the patient 0:16:01.200 --> 0:16:05.040 is going through. So like if you talk about therapists 0:16:05.040 --> 0:16:07.520 in terms of air, So there are chadbots which are 0:16:07.520 --> 0:16:10.960 coming up, like robot and new par So these chadbots 0:16:11.000 --> 0:16:14.400 that are increasingly being used to offer advice and a 0:16:14.440 --> 0:16:17.720 line of communication for mental health patients during their treatment. 0:16:18.120 --> 0:16:20.520 So they can also help with coping up with symptoms 0:16:20.560 --> 0:16:22.520 as well as they can look out for keyword that 0:16:22.560 --> 0:16:26.520 could trigger a possible help that patient needs. So chargipity 0:16:26.600 --> 0:16:28.560 can be used like a therapist. Like there have been 0:16:28.640 --> 0:16:31.280 certain use cases, like I've been reading on Reddit and 0:16:31.320 --> 0:16:33.280 there have been people who have been like sharing their 0:16:33.320 --> 0:16:36.520 stories around how they use chargity as a therapist. So 0:16:36.760 --> 0:16:39.440 when we see that chatbot can be used as a therapist, 0:16:39.720 --> 0:16:42.440 it is like we are giving them some inputs and 0:16:42.440 --> 0:16:45.080 they're basically doing sentiment analygies on the basis of textual 0:16:45.120 --> 0:16:47.440 responses that we're giving to them, and then they are 0:16:47.440 --> 0:16:50.600 basically modifying their answers to make it more human like 0:16:51.040 --> 0:16:54.880 and that's how they can work as AI therapist. But 0:16:55.080 --> 0:16:58.040 there are concerns around it as well. Like the first 0:16:58.080 --> 0:17:01.160 thing that comes up with is reliability. How much accurate 0:17:01.200 --> 0:17:04.600 of the solution that chatbot is providing us or any 0:17:04.640 --> 0:17:06.920 tool that we have built as a form of therapist 0:17:07.000 --> 0:17:11.360 is providing us. So first is reliability and then comes accountability. 0:17:11.400 --> 0:17:14.399 What if you know, something wrong happens, Who's responsible for 0:17:14.480 --> 0:17:17.679 all of this? But apart from this, the concern that 0:17:17.800 --> 0:17:20.359 always struck me is that these are privately funded apps, 0:17:20.400 --> 0:17:23.359 Like these are the apps that have been used at 0:17:23.359 --> 0:17:26.760 commercial level. I mean, there are certain subscription charges that 0:17:26.800 --> 0:17:30.120 need to be paid to use these apps. So I've 0:17:30.160 --> 0:17:33.080 always had this view that once you start commercializing and 0:17:33.119 --> 0:17:37.840 start making out profits from healthcare services, then things turn problematic, 0:17:37.920 --> 0:17:41.200 you know, and when something as vulnerable and as volatile 0:17:41.240 --> 0:17:44.199 as mental health is involved, I think we must be 0:17:44.359 --> 0:17:47.280 very much cautious. We must be very much vigilant about 0:17:47.320 --> 0:17:49.160 the kind of apps we are using and the kind 0:17:49.200 --> 0:17:51.800 of tools that are coming in in terms of mental 0:17:51.840 --> 0:17:52.879 health care services. 0:17:53.640 --> 0:17:56.000 And that leads me to if you are going to 0:17:56.000 --> 0:17:58.800 be using these sorts of chat pots like chat GPT, 0:17:58.960 --> 0:18:01.200 as you mentioned, to make sure that you're well aware 0:18:01.200 --> 0:18:04.520 of who's got your data, what the privacy concerns may be, 0:18:05.280 --> 0:18:07.919 and how you can make an informed decision. I like 0:18:07.960 --> 0:18:11.119 to get your thoughts around that, particularly around privacy and 0:18:11.200 --> 0:18:14.320 data security, and maybe you could start with how you 0:18:14.400 --> 0:18:15.760 tackled it with your app. 0:18:16.800 --> 0:18:18.960 This is one of the main concerns that come up. 0:18:19.000 --> 0:18:22.120 Like you also mentioned that whenever we are using such apps, 0:18:22.160 --> 0:18:23.720 we need to be aware that what kind of data 0:18:23.760 --> 0:18:26.520 we are feeding into it and what kind of formissions 0:18:26.520 --> 0:18:29.359 we're giving to such a tool. But someone who is 0:18:29.520 --> 0:18:32.840 going through a mental health problem mental illness, I mean 0:18:33.040 --> 0:18:36.480 we cannot say that the person is healthy enough or 0:18:36.520 --> 0:18:38.840 stable enough to be able to make a decision on this, 0:18:39.240 --> 0:18:42.840 and therefore privacy concerns will come later in the stage. 0:18:42.920 --> 0:18:44.960 But the first thing is that are we able to 0:18:45.520 --> 0:18:49.000 make the patients familiarize with the kind of data they're 0:18:49.000 --> 0:18:51.840 feeding into the apps and what are the consequences or 0:18:51.920 --> 0:18:53.919 ramifications that this data can lead to. 0:18:54.520 --> 0:18:59.080 Yeah, because I actually heard some stories around people using 0:18:59.119 --> 0:19:05.400 these chat says therapy and the concept of this transference, 0:19:05.440 --> 0:19:09.840 so they're actually falling in love with the bots. There's 0:19:09.880 --> 0:19:13.720 a similar experience with psychologists where patients fall in love 0:19:13.760 --> 0:19:17.639 with the therapist. So that's just another potential challenge that 0:19:17.680 --> 0:19:19.360 we all have to come to deal with if you're 0:19:19.359 --> 0:19:20.719 going to start using these things. 0:19:21.320 --> 0:19:24.520 Yeah, they're a virtual entities that are coming into this 0:19:24.600 --> 0:19:26.920 scenario and we are able to see them and they've 0:19:26.920 --> 0:19:30.320 been living their own life. People are becoming so comfortable 0:19:30.359 --> 0:19:33.679 with chatbirds now because definitely there's a lack of communication 0:19:33.720 --> 0:19:35.920 that is happening, and ever since the pandemic gets strung, 0:19:36.000 --> 0:19:39.199 this communication gap has increased, it has profoundly increased, so 0:19:39.240 --> 0:19:41.720 people are finding way to escape this and then these 0:19:41.760 --> 0:19:44.639 AI therapists come as a rescue and therefore people use 0:19:44.680 --> 0:19:48.560 it blindly without being enough aware about what kind of 0:19:48.640 --> 0:19:51.080 data they're feeling it and what kind of algorithms these 0:19:51.280 --> 0:19:54.760 applications are using. Because we know that to these algorithms 0:19:54.880 --> 0:19:58.320 may not be explainable, they're not transparent, so we have 0:19:58.400 --> 0:20:01.399 to be aware about this as well. Literacy and education 0:20:01.520 --> 0:20:03.320 is needed in these aspects as well. 0:20:04.040 --> 0:20:08.639 Yeah, just on that you talked about explainability and transparency, 0:20:09.240 --> 0:20:11.040 do we just explain to the audience who may be 0:20:11.119 --> 0:20:14.040 not so familiar with those terms when it comes to 0:20:14.640 --> 0:20:18.359 AI models, what that actually means transparency. 0:20:18.400 --> 0:20:20.840 Okay, So there's a term that goes with algorithms, and 0:20:20.880 --> 0:20:24.320 that's black box. So algorithms are like black blocks. We 0:20:24.400 --> 0:20:26.679 know what is going out, but we do not know how 0:20:26.800 --> 0:20:31.240 all of this is functioning, what is actually into the algorithm, 0:20:31.520 --> 0:20:33.959 and what is the procedure and how on what basis 0:20:33.960 --> 0:20:37.840 they're doing everything. Transparency is related to the kind of 0:20:37.920 --> 0:20:40.160 data we're feeding it and the way we are using 0:20:40.200 --> 0:20:43.359 it and how algorithm is working. To know this and 0:20:43.440 --> 0:20:47.520 explainability means that any user, because there are two categories 0:20:47.800 --> 0:20:51.280 of population who are associated with any AA system. The 0:20:51.359 --> 0:20:53.359 first one are users and the second one are the 0:20:53.560 --> 0:20:56.639 developers and stakeholders. So stakeholders must know that what kind 0:20:56.680 --> 0:20:59.000 of algorithm it is and there should be transparency in it. 0:20:59.160 --> 0:21:01.679 But when it comes to you user, AI systems and 0:21:01.720 --> 0:21:04.880 those algorithms must be explainable enough. I mean users are 0:21:04.880 --> 0:21:08.160 able to understand in a very human like language, that's 0:21:08.240 --> 0:21:09.960 what this algorithm is doing. 0:21:10.520 --> 0:21:14.880 That's really good And as AI emerges as this tool 0:21:14.920 --> 0:21:18.000 to help people struggling with their mental health, I'd like 0:21:18.040 --> 0:21:20.720 a few more comments just around how you see it 0:21:20.800 --> 0:21:24.280 working in tandem with the medical community to better serve 0:21:24.800 --> 0:21:28.160 their patients and their communities. Do you have any thoughts 0:21:28.240 --> 0:21:30.719 on how you know this tool can actually be used 0:21:31.359 --> 0:21:33.520 together rather than a replacement. 0:21:34.200 --> 0:21:38.280 Yeah, Basically, we always think that AI is a disruptor. 0:21:38.359 --> 0:21:41.600 We have always thought of this any technology that comes, 0:21:41.760 --> 0:21:44.040 but I've always believed that they are over here to 0:21:44.119 --> 0:21:49.000 augment our capabilities and to supplement whatever you know, roles 0:21:49.000 --> 0:21:52.600 are there. So I'm from India and the very first 0:21:52.640 --> 0:21:54.159 thing that I mean I have to cover up is 0:21:54.160 --> 0:21:57.560 that we need to educate people around mental health because 0:21:57.560 --> 0:22:01.600 in India, the most instrumental impedt in terms of mental 0:22:01.640 --> 0:22:03.960 health is lack of awareness and education. People do not 0:22:04.080 --> 0:22:06.760 know what exactly depression is, what exactly anxiety and stresses. 0:22:06.800 --> 0:22:09.320 They use it in a very casual way. And to 0:22:09.359 --> 0:22:13.120 be very honest, mental health is something which is stigmatized 0:22:13.119 --> 0:22:15.280 in India. So you know, if someone is suffering from 0:22:15.280 --> 0:22:17.600 mental health issue, they are often labeled as lunatics or 0:22:17.600 --> 0:22:20.840