Full transcript
0:04[music]
0:09[music]
0:17Hello everyone. I welcome you once again
0:19to my NPTL course introduction to
0:21medical humanities.
0:23Today we are going to talk about
0:25something which is pervading all our
0:27lives and if it is pervading all our
0:30lives, how can healthc care be
0:33different? So we talk about integrating
0:36AI into health care in the scope of
0:39modern medical humanities.
0:42Before we begin, let me give you a quick
0:44lecture overview. How am I going to talk
0:46about? I talk about infertility
0:49technology and the rise of AI in
0:52reproductive health care and even in
0:55other areas. I take up the
0:57biotechnological gaze and how it
1:00responds or corresponds to the body uh
1:03within the clinical sense and also in
1:06its social sense.
1:08A model that is very popular uh in the
1:11context of AI data feminism a very
1:15important theoretical construct. I take
1:17up that and follow it up with how
1:20algorithmic bias is leading to the
1:23politics of technology.
1:26Before I begin, let me uh take up
1:28Ishuguro's quote which says, I don't
1:32mean simply the organ. Obviously, I'm
1:35speaking in the poetic sense, the human
1:38heart. Do you think there is such a
1:40thing, something that makes each one of
1:43us special and individual?
1:46Where does it lead? This leads from the
1:48seminal question Ishiguru asked which
1:52says do you believe in human heart and
1:55if we believe in human heart in the
1:58literal sense how about the poetic
2:03when you talk about the poetic sense of
2:05the human heart you find that in the
2:10intervention of technology it is finding
2:13its meaning changed. This change
2:16indicates that with the intervention of
2:18AI, we are going to ask most of the
2:21areas where AI penetrates a fundamental
2:24question or rather fundamental questions
2:28here on your screen. The first can
2:30machines understand human emotions. So
2:33when we discuss heart not in its literal
2:36sense but in its poetic sense, can it do
2:38that? Most of the times in our daily
2:40lives we give so many analogies around
2:43heart. The metaphor of heart is used
2:46almost every day. So do you think
2:48machines can do that? Can algorithms
2:51replicate human intuition?
2:54When we talk about human intuition, we
2:57believe that each heart, even if I go by
3:00the analogy of the heart, it responds to
3:03each one of us differently. The
3:05customization is done, the
3:07personalization happens. And in the
3:10absence of that customization and
3:12personalization, can algorithmic
3:16trends replicate that? Finally, the
3:20third and important question, can data
3:23capture the complexity of human
3:26experience?
3:28Even if you compare the physical
3:30features that you can see through your
3:33naked eyes, you will find thousands and
3:36thousands of people around you very
3:39different from each other.
3:41Makes me wonder if our emotions are also
3:43different. So when we collect data, what
3:47do we do? We homogenize. So technology
3:51also in the process of being
3:53nondiscriminative
3:56runs the risk of being homogenized.
3:59Therefore these three fundamental
4:02questions you have to ask within the
4:05scope of medical humanities when you
4:08discuss AI in healthcare. Healthcare
4:12which till now in last 37 lectures we
4:15have discussed is a very specialized
4:19kind of a field which intersects in
4:22almost all the disciplines of academic
4:26study. The first area that we choose for
4:30AI intervention is infertility
4:33technology and the rise.
4:36Now there are quite a few lectures I
4:39took on the themes of reproductive
4:42justice with specific reference to how
4:46AI has intervened in the reproductive
4:48justice discourse.
4:51Here in reproductive medicine AI is
4:55seminal. If you look at your screen, you
4:58will find that AI has transformed IVF
5:02treatment in India and shown 5 to 7%
5:06increase in certain context.
5:10While it might look that it is just the
5:12increase of 2% on your screen, but if
5:15you locate the statistics on
5:19or rather in society, you will find that
5:22this is a huge number.
5:25So technology is changing the way we
5:28were perceiving reproductive medicine.
5:31When you talk of the treatments, you
5:33will see the statistics being completely
5:37different. On your screen in 2021, it
5:42showed 746 million US.
5:46By 2027, it's projected to be 1453 USD.
5:52How many IVs cycles? More than 0.5 to
5:570.6 million.
6:00Now this was just reproductive justice
6:03but even otherwise if you look at a
6:06research produced by Zeon you find that
6:09the AI in healthcare market size has
6:12increased rapidly. In 2023 it shows 0.83
6:19billion.
6:20And if you look at 2032, you find 17.75
6:2717 times more or maybe more than 17
6:30times. So quite
6:34quietly indicating the fact that AI is
6:39just everywhere
6:41including the most component of our
6:45being which is our well-being of which
6:48health is a broad indicator.
6:52When you talk of this AI spike,
6:56how does it change the discourse of
6:58medical humanities?
7:01So till now we were discussing
7:05to
7:07decode or align the biomod of medical
7:12with that of the psychosocial model of
7:15humanities.
7:18In totality this was called the
7:19biocschosocial model. It pretty much
7:23much it pretty much encompassed
7:27all aspects of medical where we
7:30discussed the narratives around medicine
7:33with that of their social implications
7:37or how they find meaning in society. How
7:40does a particular society construct the
7:44meaning of a particular
7:48physical issue?
7:50Through the narratives, through
7:52tradition, through practices, attitude,
7:55behavior,
7:57quite a lot of aspects you can deal
8:00with. But the rapid integration of AI
8:04raises questions that computer science
8:06cannot answer. And therefore there are
8:10three focal points we have to address.
8:13How we define and measure AI. How we
8:17study and experience AI and how do we
8:20safely deploy AI in clinical settings.
8:24Now with these focal points if you
8:27discuss
8:29you will find that
8:31when we think of AI there is no umbrella
8:35term for AI
8:38or AI is different for
8:43the discipline in which it is employed.
8:46Like just the other day I was uh
8:49listening to a podcast about the future
8:52of medical in India and there was a
8:57discussion on Medge Gemini.
9:01Now this is primarily for doctors.
9:04So it is different from the Gemini which
9:06is a search engine or uh a kind of an AI
9:10tool uh
9:12which is meant for doctors. So even the
9:17knowledge that the clinician has
9:22for them the AI tool is different for us
9:26the AI tool is completely altogether
9:29different. So how do we define and
9:34measure AI in a particular context is
9:39something that we should think of when
9:42we think about AI. And what does it mean
9:44in medical humanities? Not all areas of
9:47medical humanities will cater to the
9:49same kind of AI. The experience of AI
9:53will be different. So somebody who's
9:54more informed, more educated, for them
9:58AI will be different. Also, it depends
10:01on our physical capability. Now I'll
10:03tell you why I'm saying this. So I was
10:06reading a research in which
10:09it was suggested that AI is going to
10:12give medicine the work that most most of
10:14the times uh nurses do in the hospital
10:19of caregiving in terms of giving
10:21medicine that will be done by AI. Now
10:24they said that at least in the hospital
10:2680 to 90% patients are mobile so they
10:30will be able to help themselves but 10%
10:34still will be in need of human
10:36intervention. So the experience of AI in
10:41the hospital setup cannot be
10:43homogenized. It depends on the condition
10:46of the person.
10:48And then of course any kind of AI has to
10:52be
10:54ensuring the safety measures because it
10:58is related to health and any algorithm
11:01which fails can be fatal.
11:06AI is also called the third way of
11:08knowing. Now the third way of knowing
11:12one knowing which is known by the
11:15patient one knowing which is known by
11:18the doctor or clinician. So one is the
11:22patient centric and one is the clinician
11:24centric. But when AI intervenes,
11:28it's the third way of knowing. And this
11:31knowing is powerful knowing
11:34because it has data. And we know that
11:38data is power.
11:40So when something
11:43so powerful in its scope and knowledge
11:49intervenes within the human stories,
11:53they are bound to change the algorithm
11:56not just in the literal sense but in all
12:00senses possible. How do you do that? So
12:03traditionally there is a reliance on
12:06clinical judgment.
12:08You see, you touch, you get the test
12:11done all sort of things and that is
12:15supplemented by machine learning
12:17algorithms. So there is an X-ray which
12:19is done, there is an MRI, CT tests,
12:22blood tests, all kinds of things. So
12:26these tests are supportive to the
12:29clinician's sense of diagnosis. But when
12:33AI comes the clinical judgment gets
12:36algorithm decides what is the diagnosis.
12:41So you find that in reproductive
12:44technologies this is already happening
12:46where system is analyzing what fits
12:49what.
12:51But in other senses also you will find
12:54it's projected you will have hospitals
12:55and clinics and you will have virtual
12:58wards and homebased apps will connect
13:02you to those hospitals
13:05during coid9 pandemic teleconultation
13:08was something which boomed I will talk
13:10about it in the next lecture in great
13:12detail when I'm discussing future of
13:14medical humanities.
13:17Now when technology is seeping
13:22everywhere, you also find the gaze on
13:27the body
13:29beyond the clinical and social to now
13:33algorithm based.
13:36And this further fragments the body into
13:42pieces that can be read through
13:46machines.
13:48So feminist philosopher Rosie Bredotei
13:52argues that modern biosciences
13:55increasingly treat human body as a set
13:58of detachable biological components
14:01rather than a unified whole.
14:05So the body is not the body but a
14:08product. A product which is a
14:10composition of many objects together. So
14:15if we consider body as an image, it's
14:18not taken in totality. It's a collage of
14:21image that is happening. So
14:24cognition doesn't work like that. Human
14:27beings don't work like that. We work in
14:29totality. And therefore this kind of
14:32fragmentation is problematic and AI has
14:35to address. Till now we have been
14:38talking about narrative competence,
14:40cultural competence where not just the
14:43body of the person who has come with a
14:45story of medicine with a story of what
14:48he he or she is feeling but more than
14:51that we are saying no take into
14:53consideration what society thinks of
14:56that. But here now we are at the brink
15:00of
15:02further making it micro.
15:06So she says this process as the
15:09emergence of organs without bodies.
15:12So medical humanities till now has
15:16advocated for narrative and cultural
15:19competence, a patiententric model, a
15:23culture ccentric model. But machines
15:26they just concentrate on what part is
15:30affected.
15:32So the totality of it is going to miss
15:38in the context of AIdriven reproductive
15:41machine. This is all the more
15:43fragmented.
15:45And if you look at reproai which is very
15:48trending term some of you who want to
15:50work in the area of reproductive justice
15:54or who are working in the area of gender
15:58or fiction related to
16:01women
16:03or anthropology sociology
16:07such areas you will find reproi which is
16:10reproduction AI reproduction artificial
16:13intelligence but it's called repro Pro
16:15AI a multiddisciplinary technology that
16:18integrates reproductive medicine with
16:20the mathematical sciences to predict and
16:23treat infertility.
16:26This is a new kind of area which is
16:29being used in or rather utilized in the
16:34area of reproductive medicine. So when
16:37it is about more than or at least half
16:39of the population,
16:42feminist theorists have
16:45said that fragmented bodies will further
16:47fragment how women are perceived and
16:50therefore they came up with a
16:52theoretical framework called data
16:54feminism. Now data feminism was
16:57developed by Katherine Dazio and Lauren
17:01F. claim who examine how data algorithms
17:05and technologies reflect existing power
17:09and they come up with the inference that
17:13data is never neutral.
17:16And
17:19when data is produced, who is collecting
17:22data?
17:24Whose data it is?
17:27Who are people who are giving that data?
17:31from where you have collected that data.
17:35What is what are other demographic
17:37factors of the person with whom you have
17:39collected data all is important and that
17:42shapes
17:44the data that we have collected. So even
17:49if one person is missing from that data,
17:54one perspective is going for a toss.
17:58And many of you who would have studied
18:00anthropology and sociology or done any
18:03kind of data collection will say that
18:05broader the issue the larger the data
18:08sets should be. And therefore data
18:11feminist came up with seven principles
18:14framework.
18:15And what is that? Examine power.
18:19Challenge power.
18:21Rethink binaries and hierarchies.
18:24Elevate emotion and embodiment.
18:27Embrace pluralities,
18:30consider context, make labor visible.
18:33Now these were seven principles
18:37that were incorporated in data feminism
18:40and take into account the intersectional
18:42theory that I have discussed earlier
18:45which is Kimberly Krenshaw's theory of
18:48intersectionality
18:49recognizing how many other aspects of
18:54any particular society or different
18:57aspects of a particular society have to
19:00be considered while
19:03analying izing,
19:04believing and projecting data.
19:09So
19:11when you look at Sandra Harding's
19:13argument, they say that scientific
19:16knowledge is therefore produced and
19:20shaped by cultural values, institutional
19:24priorities and power relations. So the
19:27data that you collect
19:32is driven by these three factors. What
19:35is your own belief system? Where what
19:40culture do you come from? What are the
19:42kind of institutional priorities? And
19:45definitely the power equations.
19:50Therefore, you cannot help but think of
19:54algorithmic bias in the politics of
19:56technology. And Safia Umosa argues that
20:01algorithmic technologies are
20:03mathematical formulations to drive
20:05automated decisions. Now
20:09you what what does she say that you just
20:11produce that data to take up automated
20:15decisions and not have human
20:17interventions and therefore I mean Rouha
20:20Benjamin calls it gym code now gym code
20:23is the fact that in the process of
20:28technology being nondiscriminative
20:32produces or leads to more
20:34discrimination.
20:36So at all point of time technology is
20:41also driven by data and if that data is
20:44not neutral technology will be also
20:47biased. That is what they came up with.
20:50It leads to Shelley Kohhan's assertion
20:52that
20:54the concept of stratified reproduction
20:56to describe how reproductive capacities
20:59are structured through social
21:00hierarchies and AI might exerbate it if
21:04not properly regulated. So there are
21:06more regulations that are required to uh
21:10regulate AI.
21:13Quickly coming towards the end of our
21:15lecture which discusses the intervention
21:18in terms of medical humanities. The
21:21first is that in terms of boundaries it
21:24has to move from strictly technical
21:27disciplines to also human intervention.
21:31I spoke about humanizing in the last
21:34lecture. So while there are some areas
21:38of medicine that can be replaced or work
21:42in supplement to what clinicians are
21:44thinking and hence the narrative as well
21:48but at no point of time perhaps we can
21:51do away with the human intervention
21:53because this is about human beings and
21:55it is about their life.
21:58In terms of methodology, the STS uh
22:01science, technology, studies and
22:03humanities tools navigate around social
22:07and cultural shift because care moves
22:10from clinics to apps and virtual worlds.
22:13So when the caregiving is not having any
22:18human input, how does one feel? So
22:21technology is
22:24going to struggle with this aspect also
22:28because well-being till now we have been
22:31trained to believe that it is
22:33humanentric
22:34and therefore they have to have
22:38professional guidelines where the center
22:41is situated and lived experience of
22:44diverse patients to build ethical
22:48effective framework for clinics.
22:51and how they deploy AI is suggested.
22:56Thank you so much. I will in the next
22:58lecture take up future of medical
23:01humanities and before we summarize the
23:04entire course in lecture 40 that will be
23:06my last lecture. We have almost reached
23:10uh the end. Uh I hope uh you go through
23:14the entire slides all over again and
23:17hopefully I can summarize them in the
23:20lecture 40. Thank you so much.