1
00:00:02,080 --> 00:00:10,450
Thank you very much, thank you for the introduction. It's going to be a bit of a change in shifts in terms of the scale of the project I'm working on.

2
00:00:10,450 --> 00:00:16,810
That's the title of my project and I should say it's been funded by my senior fellowship, to my own debates and their fellowship.

3
00:00:16,810 --> 00:00:19,750
And it's a project that is a bit more theoretical in some ways.

4
00:00:19,750 --> 00:00:25,990
I again, to the one that that is, and although there is a strong theoretical component, and I think there is also overlap,

5
00:00:25,990 --> 00:00:33,400
as you would say, between some particular kind of principles, health principles and notions that are used in my project as well.

6
00:00:33,400 --> 00:00:39,640
So I would like to start with just giving a definition of what I mean by a I just to make sure that we're all on the same page.

7
00:00:39,640 --> 00:00:49,270
So I think I mean, computational systems used to resolve problems that would normally require human intelligent input to be resolved.

8
00:00:49,270 --> 00:00:57,790
And this is a technology that has been heralded as kind of bringing about a new, exciting revolution in health care.

9
00:00:57,790 --> 00:01:03,670
So A.I. has a great potential to improve health outcomes, as we've heard, as well as the individual,

10
00:01:03,670 --> 00:01:08,050
but also the population level assist cash-strapped health care systems like the

11
00:01:08,050 --> 00:01:12,910
NHS at the moment by significantly increasing efficiency and effectiveness

12
00:01:12,910 --> 00:01:23,340
in the way that we use resources and generally improve access to health care and health care services for the population and individuals as well.

13
00:01:23,340 --> 00:01:28,350
So some of the ways in which I can have a positive impact, particularly on population health,

14
00:01:28,350 --> 00:01:32,460
are through, for example, more effective prevention methods,

15
00:01:32,460 --> 00:01:40,140
providing more accurate early diagnosis of conditions, better monitoring of infectious diseases and assisting with epidemic prediction.

16
00:01:40,140 --> 00:01:49,350
We're going through the coronavirus epidemic at the moment, so air is one of these things that could definitely help kerb the spread of the disease.

17
00:01:49,350 --> 00:01:53,050
Also through personalised promotion of healthier living.

18
00:01:53,050 --> 00:01:57,870
So the thing about apps that they can give you advice on what to eat and how

19
00:01:57,870 --> 00:02:03,600
much to exercise and not exercise based on your particular health risk profile.

20
00:02:03,600 --> 00:02:12,540
Also, air tools can be used in self-management of chronic diseases like diabetes or certain types of cardiovascular diseases,

21
00:02:12,540 --> 00:02:18,810
and also improve the efficiency of patient management on the ground for us to assess health care systems.

22
00:02:18,810 --> 00:02:24,300
We very often hear in the news how the waiting time in A&E are getting longer and longer,

23
00:02:24,300 --> 00:02:28,740
and patients cancer patients cannot access care as fast as they should,

24
00:02:28,740 --> 00:02:37,970
and there are already items developed and implemented in hospitals around the country to assist in dealing with these type of problems.

25
00:02:37,970 --> 00:02:42,470
And of course, it's not the first time that the health care and health systems are turning to technologies

26
00:02:42,470 --> 00:02:47,270
to kind of help them deal with and improve health outcomes and deal with problems as well.

27
00:02:47,270 --> 00:02:48,650
I mean, think of the stethoscope.

28
00:02:48,650 --> 00:02:56,330
It was one of these kind of first technology documents the hearing of healthcare professionals that could provide much more accurate diagnosis,

29
00:02:56,330 --> 00:03:04,520
that they could hear what was going on. Same with X-rays augmented our capacity to see inside the body without having to open the body up.

30
00:03:04,520 --> 00:03:08,030
So it's as I said, it's not the first time that we have technologies coming in.

31
00:03:08,030 --> 00:03:14,600
But I believe that air is different because the aim of AI is not to augment our physical capabilities,

32
00:03:14,600 --> 00:03:21,530
but in a way it's trying to augment our rational decision making capacity and lead to rational action.

33
00:03:21,530 --> 00:03:25,700
And it is this understanding of rationality, I think,

34
00:03:25,700 --> 00:03:31,560
as being the ability to compute solutions that would optimise certain values given

35
00:03:31,560 --> 00:03:37,020
the available information that we have is what makes a different type of technology.

36
00:03:37,020 --> 00:03:46,880
I think that Stuart Russell called that calculative rationality, and I think it's one that can be truly disruptive in this area of health care.

37
00:03:46,880 --> 00:03:49,400
And what I mean by disruptive.

38
00:03:49,400 --> 00:03:56,900
What I mean is that it can fundamentally change our expectations and also our perception of what health care is about, what is trying to achieve.

39
00:03:56,900 --> 00:04:02,110
Also the role of healthcare professionals and the role of systems within that context.

40
00:04:02,110 --> 00:04:10,880
And with this in mind, I would like to turn to these three notions of I'm kind of particularly interesting theme of efficiency, accuracy and trust.

41
00:04:10,880 --> 00:04:18,710
So as we try to draw from cloud as well and make before efficiency and accuracy are obviously clearly important for patients,

42
00:04:18,710 --> 00:04:22,820
for health care professionals and for the health care system as a whole.

43
00:04:22,820 --> 00:04:27,950
Having access to more accurate diagnoses or more accurate information about the spread of infectious

44
00:04:27,950 --> 00:04:34,790
diseases means a much increased probability to have a positive outcome and effectively save lives.

45
00:04:34,790 --> 00:04:39,680
Equally, greater efficiency and effectiveness in the way that we use our resources.

46
00:04:39,680 --> 00:04:45,710
Time, like the time of health care professionals financial resources means waste avoidance and my ability to

47
00:04:45,710 --> 00:04:53,990
channel effort to the right place in the right way at the right time and again improve health outcomes.

48
00:04:53,990 --> 00:04:59,030
But what so what we're kind of theoretically facing in front of us and maybe, you know,

49
00:04:59,030 --> 00:05:03,770
maybe it is already kind of happening out there as some of the examples are coming already

50
00:05:03,770 --> 00:05:09,200
through is a kind of win win situation where patients benefit from more accurate diagnoses,

51
00:05:09,200 --> 00:05:18,620
targeted screening prevention, better outcomes and health systems, increase efficiency, reduce costs and keep populations have that for longer.

52
00:05:18,620 --> 00:05:20,630
However, it is currently, I think,

53
00:05:20,630 --> 00:05:28,070
not quite certain whether the incorporation of A.I. into population health and healthcare systems in general will be as frictionless and

54
00:05:28,070 --> 00:05:35,840
as unproblematic as we would like to hope that it will be and how to address potential ethical issues that might arise in the context.

55
00:05:35,840 --> 00:05:43,520
And I think one of the issues that is particularly interesting and important that we need to address in this context is the issue of trust.

56
00:05:43,520 --> 00:05:47,600
So the way I understand trust is like trust is a fundamental value of health care

57
00:05:47,600 --> 00:05:52,340
provision and underpins the relationship between patients and the health care system.

58
00:05:52,340 --> 00:05:58,820
Patients trust health care professionals and the system as a whole because they believe that people working in it have the right skills,

59
00:05:58,820 --> 00:06:04,520
the right knowledge, the right expertise and also more latitude to care for them.

60
00:06:04,520 --> 00:06:09,020
So we often hear how the NHS is the most trusted institution in the UK.

61
00:06:09,020 --> 00:06:14,090
So what characterises trust is an acceptance of and of uncertainty from the position of

62
00:06:14,090 --> 00:06:18,500
the patient and the position of vulnerability that the patients put themselves into,

63
00:06:18,500 --> 00:06:23,990
which is, however, justified because we have this belief that the people who are caring for us have the right skills,

64
00:06:23,990 --> 00:06:29,270
but also the right attitude towards us that they are, they have a good disposition towards us.

65
00:06:29,270 --> 00:06:36,710
So the question then becomes like if we start outsourcing clinical expertise or the skin side of trust to machines,

66
00:06:36,710 --> 00:06:42,590
would that challenge this trust and alter perceptions regarding health care professionals and their role within the system?

67
00:06:42,590 --> 00:06:44,730
So it is a question of whether patients,

68
00:06:44,730 --> 00:06:54,240
once they consider doctors and nurses as trustworthy professionals whose recommendations they ought to follow when when they went to see the.

69
00:06:54,240 --> 00:07:01,180
But trust is also important in the relationship that health care professionals develop with technologies and that technology providers as well.

70
00:07:01,180 --> 00:07:07,170
Companies developing a I already know that the consumer trust is crucial for the acceptability of that product.

71
00:07:07,170 --> 00:07:15,330
It is particularly important in care provision where the health care system has a fiduciary obligation for patients to always act that their best

72
00:07:15,330 --> 00:07:23,550
interests are what the doctors and nurses be confident to prescribe a self-monitoring choice to patients if they do not understand how they work.

73
00:07:23,550 --> 00:07:28,290
How should this agreement between health care professionals and A.I. systems be resolved?

74
00:07:28,290 --> 00:07:35,220
Should doctors override air produce recommendations or not? I think probably you are familiar.

75
00:07:35,220 --> 00:07:40,110
We are back with a IBM Watson Watson case.

76
00:07:40,110 --> 00:07:48,840
I think 2018 that developed a cancer diagnosis tool that was introduced in many hospitals to some hospitals in the US and within,

77
00:07:48,840 --> 00:07:52,560
I think within the year or less than that, it was taken down.

78
00:07:52,560 --> 00:08:00,120
And the main reason was that the healthcare professionals do not trust the tool, so they wouldn't use it.

79
00:08:00,120 --> 00:08:04,650
However, I mean, it is true that as A.I. develops and improves, I mean, the hope,

80
00:08:04,650 --> 00:08:10,170
at least is that and as the accuracy of prediction above gets better and better,

81
00:08:10,170 --> 00:08:19,260
and these tools improve the accuracy that these tools will kind of replace the need for for trust.

82
00:08:19,260 --> 00:08:27,030
Some some theory is kind of suggesting that. So the hope is that their risk profile, introduction and diagnosis becomes more accurate.

83
00:08:27,030 --> 00:08:33,060
With the use of A.I. in big data, healthcare professional patients and the public will not need to trust anymore by

84
00:08:33,060 --> 00:08:38,190
the will and instead to rely on intelligent machines to guide health behaviour.

85
00:08:38,190 --> 00:08:45,300
So the likely effects of a health care system that is reliable but not necessarily trusted or trustworthy are unclear.

86
00:08:45,300 --> 00:08:52,680
Could it improve health outcomes and save more lives? Or that lead to the reinforcement of more reductionist view of health care,

87
00:08:52,680 --> 00:08:58,020
which has been so far associated with poor health outcomes and patient outcomes?

88
00:08:58,020 --> 00:09:07,170
And beyond this kind of practical concerns is something intangible yet more important lost if trust becomes obsolete in the health care system.

89
00:09:07,170 --> 00:09:10,560
So these are some of the questions that my project is trying to.

90
00:09:10,560 --> 00:09:13,110
I'm trying to answer. So I would like to explore.

91
00:09:13,110 --> 00:09:22,470
I'm starting to explore this relationship between efficiency, accuracy and trust in an AI augmented population, health and health care.

92
00:09:22,470 --> 00:09:28,020
So questions like what is the value and meaning of these notions in this new in this context?

93
00:09:28,020 --> 00:09:32,370
Should we start shifting our understanding of what is required in health?

94
00:09:32,370 --> 00:09:40,500
Or should we try to steer technology to accept some and incorporate certain established values into their system as well?

95
00:09:40,500 --> 00:09:44,310
And I think these are both practical questions, but also importantly,

96
00:09:44,310 --> 00:09:49,770
ethical questions that we will need to resolve as we move forward with with AI in healthcare.

97
00:09:49,770 --> 00:09:54,048
Thank you very much.