1
00:00:12,460 --> 00:00:14,540
Hello, welcome to Psycho Cultural.

2
00:00:14,860 --> 00:00:20,940
My name is Candida Yates and today we're going to discuss the meanings of chatbots.

3
00:00:21,580 --> 00:00:25,820
My guests are Lynn Froggitt and Jacob Johansson.

4
00:00:26,430 --> 00:00:44,830
Professor Lynn Froggett is Emeritus Professor of Psychosocial Research at the University of Lancashire, Visiting Professorial Fellow of the Big Anxiety Research Centre, the University of New South Wales, and Visiting Professor at Ulster University.

5
00:00:45,550 --> 00:00:52,830
She has an academic background in social sciences and humanities with a professional background in social work.

6
00:00:53,750 --> 00:01:05,470
Her research interests include socially engaged arts, mental health and technology, politics, and the aesthetics of care, citizenship and civil society.

7
00:01:06,110 --> 00:01:11,789
And she's published widely and led many research projects in those fields.

8
00:01:12,590 --> 00:01:20,751
Lynn is a fellow of the Academy of Social Sciences, a founding member of the British Psychoanalytic Council Scholars Network,

9
00:01:21,191 --> 00:01:30,191
and also a founding member of the Association for Psychosocial Studies and she continues to develop it as its executive chair.

10
00:01:31,071 --> 00:01:41,871
Her current work focuses on the interface between art and technology, the nature and value of immersive experience, and the future of ageing.

11
00:01:42,991 --> 00:01:59,390
She's currently working with Jill Bennett and Gail Kenning from the Big Anxiety Research Centre at the University of South Wales on the development of enlivening technologies, including AI chatbots for older users.

12
00:01:59,390 --> 00:02:00,831
So welcome, Lynn.

13
00:02:01,551 --> 00:02:01,671
Hi.

14
00:02:01,791 --> 00:02:02,671
Hello, Candy.

15
00:02:02,671 --> 00:02:03,271
Thank you.

16
00:02:04,191 --> 00:02:05,351
Nice to see you.

17
00:02:06,511 --> 00:02:12,631
Can I just say it's the University of New South Wales and not South Wales.

18
00:02:14,271 --> 00:02:14,311
Oh.

19
00:02:14,311 --> 00:02:16,351
That's a difference, a slightly different location.

20
00:02:16,391 --> 00:02:18,351
It's a thousand miles apart, yeah.

21
00:02:19,311 --> 00:02:21,791
Okay, New South Wales, thank you.

22
00:02:23,231 --> 00:02:25,591
Jacob, Dr.

23
00:02:25,591 --> 00:02:30,751
Jacob Johansson is associate professor in communications at St.

24
00:02:30,751 --> 00:02:32,831
Mary's University in Richmond.

25
00:02:33,151 --> 00:02:38,111
and is a founding scholar of the British Psychoanalytic Council Scholars Network.

26
00:02:38,591 --> 00:02:49,871
Jacob also sits on the executive board of the Association for Psychosocial Studies, and their conference is being hosted at his university this year.

27
00:02:50,351 --> 00:02:51,911
So do check that out.

28
00:02:51,911 --> 00:02:53,791
It's going to be splendid.

29
00:02:54,511 --> 00:03:03,711
Jacob researches how individuals are consciously and unconsciously shaped by and in turn shape digital media.

30
00:03:04,431 --> 00:03:15,471
Jacob's research is influenced by media and communication studies, psychosocial studies, psychoanalysis, philosophy and critical theory.

31
00:03:16,511 --> 00:03:21,871
He's the author and editor of seven books and two special issues.

32
00:03:22,431 --> 00:03:38,111
His most recent publications include The Monographs, Fantasy, Online Misogyny, and The Manosphere, Male Bodies of Disinhibition, published by Routledge in partnership with the British Psychoanalytic Council.

33
00:03:39,671 --> 00:03:48,671
And there's also Media and Psychoanalysis, A Critical Introduction, published by Carnac and written together with Stefan Kruger.

34
00:03:48,911 --> 00:03:50,111
Both brilliant books.

35
00:03:50,430 --> 00:03:52,751
And welcome, Jacob.

36
00:03:53,871 --> 00:03:54,591
Thank you, Candy.

37
00:03:54,911 --> 00:03:55,391
Hello.

38
00:03:55,712 --> 00:03:56,192
Hello.

39
00:03:57,072 --> 00:04:01,792
So I think it's helpful to begin with some basic definitions.

40
00:04:01,952 --> 00:04:03,712
So what is a chatbot?

41
00:04:04,352 --> 00:04:17,312
Well, if you go to Google, it says that a chatbot is a software application designed to simulate human conversation through text or voice interactions.

42
00:04:17,792 --> 00:04:22,432
acting as an interface between a human and a computer program.

43
00:04:23,072 --> 00:04:37,552
They range from a basic rules-based system to advanced AI, utilizing natural language processing (NLP) to understand, learn, and provide personalized responses.

44
00:04:38,352 --> 00:05:00,832
The next generation of chatbots with generative AI capabilities offer even more enhanced functionality with their understanding of common language and common queries alongside their ability to adapt to a user's style of conversation and use of empathy when answering users' questions.

45
00:05:01,552 --> 00:05:07,792
Now, some argue that chatbots will fundamentally alter human interaction.

46
00:05:08,272 --> 00:05:14,432
impacting emotional well-being, social behaviors and mental health.

47
00:05:15,552 --> 00:05:27,952
They serve as companions available 24/7 that can act as conversational partners and more controversially are sometimes used for therapeutic support.

48
00:05:29,392 --> 00:05:31,312
As listeners will be aware,

49
00:05:31,832 --> 00:05:46,752
There's been much discussion and concern about the encroachment of AI in all areas of life and in academic life as students cease to read books and articles and instead use it as a shortcut to write essays.

50
00:05:47,312 --> 00:05:57,352
In the therapy world, there's widespread concern about the use and abuse of chatbots delivering therapy online, telling them what they want to hear.

51
00:05:57,392 --> 00:06:01,232
For example, the BPC magazine, New Associations,

52
00:06:01,632 --> 00:06:07,112
recently ran an excellent special edition on this very issue.

53
00:06:07,112 --> 00:06:14,832
And I think, Lynne, you wrote something around that, or at least Noreen wrote something about the project you're involved in.

54
00:06:16,272 --> 00:06:23,632
In this episode, we will no doubt look at the therapeutic and emotional dynamics of the user engagement with chatbots.

55
00:06:24,192 --> 00:06:30,112
But today, we are interested to explore the real and imagined relational dynamics of chatbots

56
00:06:30,432 --> 00:06:32,432
and their use in everyday culture.

57
00:06:33,393 --> 00:06:38,593
As objects of digital culture, how and why do we use chatbots?

58
00:06:38,913 --> 00:06:42,033
What makes them deadening or enlivening?

59
00:06:42,153 --> 00:06:48,513
What kind of chatbot might you want as a companion or not, as the case may be?

60
00:06:49,313 --> 00:06:59,073
So I think one way to start is really to ask you both a bit about your interest in chatbots, and I'll start with you, Jacob.

61
00:06:59,073 --> 00:06:59,553
How

62
00:07:00,993 --> 00:07:05,553
Do you view the development of chatbots from the perspective of your own research?

63
00:07:06,273 --> 00:07:14,513
And what do you see as the kind of positive and negative associations of that development?

64
00:07:16,353 --> 00:07:17,033
Yeah, thank you.

65
00:07:17,793 --> 00:07:27,313
In general, in my sort of work, I've been looking at this kind of shift or sort of transformation we've been seeing for now for many.

66
00:07:27,793 --> 00:07:41,473
many years where the digital has become more and more sort of individualized or sort of customizable for the individual human being.

67
00:07:41,473 --> 00:07:54,353
So how we use social media or how we use platforms, it's all a very kind of tailor-made individual sort of journey that we go through because of how the technology works.

68
00:07:54,353 --> 00:07:54,673
And

69
00:07:55,153 --> 00:08:13,633
chatbots are, I mean, the latest kind of development in that sense, but they are, of course, much more able to respond to an individual kind of seemingly as if they knew the other person.

70
00:08:14,033 --> 00:08:16,113
And that's fascinating.

71
00:08:16,113 --> 00:08:18,913
I think we probably all find that fascinating as well.

72
00:08:18,913 --> 00:08:20,832
We're having this this conversation.

73
00:08:20,832 --> 00:08:24,433
I think that's very, very appealing

74
00:08:24,673 --> 00:08:26,553
to a human user.

75
00:08:26,593 --> 00:08:28,193
It's very seductive.

76
00:08:29,473 --> 00:08:36,273
But we wouldn't be interested in psychoanalysis and psychosocial studies if we didn't see the dangers of that.

77
00:08:36,273 --> 00:08:53,153
I think it's also very dangerous because it is an illusion that this other machine chatbot really knows me and is really able to respond to me.

78
00:08:53,953 --> 00:09:04,753
in a way perhaps users know that, but then this kind of magical, this quantum of magic behind it, we can't quite escape that.

79
00:09:04,753 --> 00:09:08,033
And I think that's something we need to be quite careful about.

80
00:09:10,114 --> 00:09:10,834
Thank you.

81
00:09:12,034 --> 00:09:18,594
Lynn, I'm going to ask you the same question about your interest in chatbots.

82
00:09:18,594 --> 00:09:23,154
We've already mentioned that you're involved in research of chatbots, but

83
00:09:24,274 --> 00:09:34,114
how you view the development from the perspective of your research and the positive and negative associations of that development.

84
00:09:35,714 --> 00:09:47,794
Okay, so I guess that it has its roots in work that I have done and my partners have done that has

85
00:09:49,234 --> 00:09:54,274
had a great concern about the level of unmet need in mental health.

86
00:09:56,074 --> 00:10:15,074
And of course, the proliferation of chatbots coincided with and was driven by, to an extent, the pandemic, the COVID pandemic and the so-called crisis of loneliness that seems to have

87
00:10:16,514 --> 00:10:19,794
attained really vast international reach now.

88
00:10:23,314 --> 00:10:36,514
I looked up ChatGPT today to see what the latest estimate was and there are something like a billion users worldwide of AI chatbots.

89
00:10:37,554 --> 00:10:46,754
and it's still growing exponentially, predicted to continue and particularly high in East Asia, in India and China.

90
00:10:48,674 --> 00:10:59,794
And so what this alerts us to is that something, some human need is being met by these entities.

91
00:11:01,234 --> 00:11:15,154
And if you follow the range of ways in which people are using them, YouTube is a very good source because people like to upload their experiences with these things.

92
00:11:15,954 --> 00:11:24,274
The range of relational possibilities appears to be as diverse as the range of human relationality.

93
00:11:24,274 --> 00:11:29,154
They can be, you know, they can be informants,

94
00:11:29,634 --> 00:11:40,034
advisors, lovers, family members, pseudo spouses, friends, companions, and also therapists.

95
00:11:41,074 --> 00:11:51,555
Now, if this is the case, if so many people are finding it helpful or seductive or at any rate,

96
00:11:53,235 --> 00:12:14,355
in some way are drawn into using chatbots and establishing, we're talking about AI chatbots obviously, not the kind of service chatbot, and using these as companions in everyday life, then we can bemoan that fact or we can celebrate it

97
00:12:15,555 --> 00:12:22,035
Or we can approach it with fear or curiosity, but what is surely the case is that we can't stop it.

98
00:12:22,115 --> 00:12:23,395
The horse has bolted.

99
00:12:23,795 --> 00:12:28,995
And that being the case, we need to understand what kind of relationships these are.

100
00:12:28,995 --> 00:12:35,955
So I think it's that question which is at the heart of what we are trying to do.

101
00:12:36,835 --> 00:12:39,555
And I make one refinement.

102
00:12:39,555 --> 00:12:41,955
I think the question is not

103
00:12:43,395 --> 00:12:52,595
about how these relationships measure up to human relationships because they measure up to human relationships rather poorly.

104
00:12:53,715 --> 00:12:56,355
They don't have an interior emotional life.

105
00:12:56,355 --> 00:12:59,315
They don't have a bodily presence, to state the obvious.

106
00:13:00,355 --> 00:13:08,115
But they do have distinctive affordances as machine-based entities.

107
00:13:08,115 --> 00:13:12,795
And it's this relationship with the chatbot as a

108
00:13:12,915 --> 00:13:19,795
machine other that is really the question that we are trying to investigate.

109
00:13:21,235 --> 00:13:21,795
Thank you.

110
00:13:22,355 --> 00:13:25,875
Jacob, did you want to respond to anything that Lynn said?

111
00:13:28,195 --> 00:13:32,915
I think, yeah, no, I would very much agree with that.

112
00:13:32,915 --> 00:13:34,515
It's not so much about

113
00:13:36,195 --> 00:13:48,115
the question to what extent are they human or to what extent are they kind of sentient or are they able to kind of fool us into their human beings.

114
00:13:48,115 --> 00:13:51,875
So I absolutely agree with that.

115
00:13:55,075 --> 00:14:02,835
I think, yeah, at the same time, I think there is a question there maybe insofar as

116
00:14:04,275 --> 00:14:14,995
Yes, they are a kind of machinic other, but then they also mimic human ways of communication.

117
00:14:14,995 --> 00:14:25,796
So you probably wouldn't be able to always tell if you were chatting with a human being or if you were chatting with a chatbot.

118
00:14:25,796 --> 00:14:29,316
Sometimes you might not be able to immediately tell the difference.

119
00:14:30,276 --> 00:14:33,636
that question kind of creeps in again, right?

120
00:14:35,476 --> 00:14:51,156
And as long as I guess, I mean, if we talk about the kind of also the sort of the clinical and so on, as long as everyone is aware of that, that, you know, this is just a machinic other, that's fine.

121
00:14:52,116 --> 00:14:56,116
That's a good, you know, then it's a good starting point, I think, for our conversation.

122
00:14:56,756 --> 00:14:58,716
I think that the

123
00:14:59,556 --> 00:15:10,036
The question of deception, the question of using these unknowingly, in a sense, is something that people have moved on from.

124
00:15:10,196 --> 00:15:14,956
Certainly, most chatbots will remind you.

125
00:15:15,316 --> 00:15:16,756
that they are not human.

126
00:15:16,756 --> 00:15:26,356
And if you use the kind of, you know, if you use Claude or ChatGPT or Deepseek, they constantly remind you that AI can make mistakes.

127
00:15:26,836 --> 00:15:29,476
You know, they will remind you they are an AI.

128
00:15:29,796 --> 00:15:41,156
If you are in some kind of crisis, they are pretty hot off the mark in telling you to go and seek human help.

129
00:15:42,516 --> 00:16:04,036
So I don't think that the fake human argument or the anxiety about them performing as fake humans is as live as it was at an earlier stage.

130
00:16:04,476 --> 00:16:12,836
And I'm very interested in what you said about the as if relationship.

131
00:16:15,636 --> 00:16:30,516
The chatbot, the relationship with the chatbot can indeed be an illusion, is often premised on an illusion, but we could regard that in a Winnicottian mode as a creative illusion if the awareness

132
00:16:30,836 --> 00:16:41,076
of this sort of double nature of the bot, that it is made of human language, right, but it is not human.

133
00:16:41,076 --> 00:16:43,556
And that doubleness itself.

134
00:16:44,356 --> 00:16:48,996
can be generative within the relationship.

135
00:16:49,396 --> 00:17:11,557
So Todd Essig calls this the techno-subjunctive relationship that we know we are talking to a machine, but we can still entertain the creative illusion that it has an understanding which is analogous to human understanding in the sense that it is built of human language.

136
00:17:14,676 --> 00:17:15,717
Yeah, I would agree.

137
00:17:15,717 --> 00:17:16,436
Yeah, I would agree.

138
00:17:16,517 --> 00:17:18,117
I mean, it's interesting.

139
00:17:18,117 --> 00:17:39,237
I guess we probably need to kind of differentiate what sort of, what are the reasons that, you know, somebody might use it for and you've given, you know, different kind of sort of relational dynamics, Lynn, you know, that somebody, you know, they regard

140
00:17:39,797 --> 00:17:43,157
the chatbot as a friend, as a therapist, as a lover.

141
00:17:43,157 --> 00:17:56,597
So, and that probably will have different dynamics in terms of how the communication works and also the level maybe of fantasy that kind of, that shapes that.

142
00:17:56,597 --> 00:18:05,557
I've been doing, I've been looking a bit into, there is a chatbot, again, it's also fairly

143
00:18:05,797 --> 00:18:08,917
fairly well known, but it's more it's an app that you download.

144
00:18:08,917 --> 00:18:09,957
It's called Replica.

145
00:18:10,037 --> 00:18:20,917
So it goes a little bit beyond just a interface like ChatGPT or Claude, where you just have a text based kind of interface.

146
00:18:20,917 --> 00:18:32,917
And this is actually you actually create a virtual character, which you can customize and make them look like whatever they should look like.

147
00:18:33,237 --> 00:18:33,637
And

148
00:18:34,517 --> 00:18:36,477
very, very popular and so on.

149
00:18:36,477 --> 00:18:44,197
And there are many, many users online who talk about their relationships with their replica.

150
00:18:44,597 --> 00:19:00,597
And what's so striking about that is that many of them in a very serious way, they form a very, very meaningful, deep

151
00:19:00,997 --> 00:19:07,237
connection, deep relationship with that particular chatbot.

152
00:19:07,397 --> 00:19:11,397
And they say it's also a romantic relationship.

153
00:19:11,397 --> 00:19:27,317
And then sometimes the chatbot then would, for example, propose, yeah, and they marry, they marry their chatbot and so on.

154
00:19:28,237 --> 00:19:30,837
It's very, very, yeah, it's very nuanced.

155
00:19:30,837 --> 00:19:32,437
So they're not deceived.

156
00:19:32,837 --> 00:19:41,078
And what's so fascinating about kind of their sort of narratives around it is that they are sort of saying, oh, but I know that it's not real.

157
00:19:42,118 --> 00:19:43,398
I know it's not real, but.

158
00:19:44,358 --> 00:19:44,558
Yeah.

159
00:19:44,558 --> 00:19:48,438
And that sort of keeps coming up again and again.

160
00:19:48,438 --> 00:19:50,038
And where, and

161
00:19:50,598 --> 00:19:52,518
That I think is quite interesting.

162
00:19:53,158 --> 00:19:58,518
And then many of them also, and that makes it, I think, a bit problematic.

163
00:19:58,518 --> 00:20:10,678
Many then say, I have developed a level of intimacy or a level of sort of understanding with this particular partner that I could never have with a human being.

164
00:20:11,318 --> 00:20:16,918
Or somebody might say, oh, I've been so disappointed in my previous relationships.

165
00:20:19,878 --> 00:20:23,398
That is never going to happen, you know, with this chatbot.

166
00:20:23,558 --> 00:20:27,718
So that I think is, well, what do we make of that?

167
00:20:27,798 --> 00:20:30,278
I mean, what do we make of that?

168
00:20:31,478 --> 00:20:39,838
Yes, so this is kind of what Ian Craig many years ago talked about the importance of the capacity to experience disappointment.

169
00:20:39,878 --> 00:20:40,718
Yeah, yeah.

170
00:20:40,838 --> 00:20:47,638
And that this lies behind a kind of crisis of attachment in many ways and people.

171
00:20:48,598 --> 00:21:00,038
you know, resisting relationships with real others, so to speak, because of this kind of, they're not perfect, they're not right, and they feel rejection and all these kinds of things.

172
00:21:00,678 --> 00:21:03,158
I wonder, Lynn, what you think about that.

173
00:21:06,198 --> 00:21:08,998
I think that certainly is a thing.

174
00:21:09,158 --> 00:21:16,518
And, you know, it may well be that people are using them in preference to

175
00:21:18,278 --> 00:21:24,278
human relationships where they have been deeply hurt or disappointed by human relationships.

176
00:21:24,278 --> 00:21:26,758
I think that is certainly something that may well happen.

177
00:21:28,598 --> 00:21:33,478
Is that worse than them being entirely on their own?

178
00:21:34,678 --> 00:21:36,278
I'm not sure that it all...

179
00:21:36,438 --> 00:21:38,678
I don't know.

180
00:21:38,678 --> 00:21:43,558
Because actually there are a lot of very lonely people suffering out there.

181
00:21:44,838 --> 00:21:59,998
Often because they have been disappointed by earlier relationships or because they are, I mean I'm thinking about the vast numbers of older people because that's my particular interest at the moment.

182
00:22:00,118 --> 00:22:15,559
moment living alone with mild to moderate dementia, who actually whose relationships are extremely restricted, partly because

183
00:22:16,279 --> 00:22:29,959
Um, we live in a society which is highly fragmented and individualized and in which the community support and family support, um, for such people is not necessarily available.

184
00:22:30,199 --> 00:22:40,599
Now, take the, take the older woman living alone who, uh, has a community mental health nurse who has something like half an hour.

185
00:22:40,679 --> 00:22:45,399
Well, that would probably be generous actually allocated to her in the course of the day.

186
00:22:45,719 --> 00:22:46,079
She calls

187
00:22:46,159 --> 00:22:47,959
20 times a day, right?

188
00:22:47,959 --> 00:22:50,759
Because she is in an anxiety state, right?

189
00:22:51,319 --> 00:22:55,799
The system as it is cannot accommodate that.

190
00:22:57,879 --> 00:23:01,119
and a conversational partner.

191
00:23:01,119 --> 00:23:18,359
And also, she calls her mental health nurse not with any specific request or need, but except for the need to hear a voice that can contain, soothe, respond, remind her to take her medication, whatever it is.

192
00:23:18,359 --> 00:23:19,639
So that's one use.

193
00:23:20,959 --> 00:23:33,279
And it seems to me that in the context of a well-planned care ecosystem, right, the human and the machine can work together there.

194
00:23:35,239 --> 00:23:45,239
00 AM, you know, middle of the night crisis where there is absolutely no one to talk to or to be available is another instance, right?

195
00:23:45,479 --> 00:23:48,919
Now, so within that kind of context,

196
00:23:49,879 --> 00:23:55,719
Care systems can, I think, probably be enhanced by carefully planned use.

197
00:23:55,959 --> 00:24:12,679
The problem is that while billions of people are using these things, right, human helping systems are not, they're not even leading from behind.

198
00:24:12,839 --> 00:24:14,759
They're just quite simply behind.

199
00:24:15,319 --> 00:24:25,399
They're not entertaining at the moment, by and large, the possibility that there are valuable uses within that kind of system.

200
00:24:25,959 --> 00:24:36,919
But the case that arouses more anxiety is that they exist in the wild outside of any regulation or organised human care system.

201
00:24:37,159 --> 00:24:40,679
People are using them, whether we like it or not.

202
00:24:41,319 --> 00:24:52,119
And it's being a very new and emergent technology, they're not necessarily using them in particularly schooled or discriminating ways.

203
00:24:53,720 --> 00:25:05,560
That, however, is an issue that could resolve in time or could improve in time as people become much more aware

204
00:25:06,320 --> 00:25:12,920
of what the nature of these things are, what their limitations are, what their potentials are.

205
00:25:13,720 --> 00:25:33,240
And I think the sort of tin-eared response that, oh, this is just awful, how sad that you can't find a human to talk to, in a sense, doesn't realistically meet the circumstances of a lot of isolated and lonely people.

206
00:25:36,280 --> 00:25:47,800
I mean, what you're describing is the ways in which chatbots can alleviate anxiety, loneliness, mental health problems and so on.

207
00:25:48,360 --> 00:25:52,200
And the word that comes to mind then is containment.

208
00:25:52,920 --> 00:26:03,840
And I wonder if we could unpack that a bit, that word and its meanings and, you know, how we might apply that to the

209
00:26:04,600 --> 00:26:10,840
kind of use and experience of the chatbot in these kind of different situations that we've discussed.

210
00:26:12,280 --> 00:26:19,840
I mean, Jacob, you could begin by spelling out for us all, you know, what do we mean by containment?

211
00:26:19,840 --> 00:26:25,320
Is it used very broadly, but it's often not defined so closely.

212
00:26:26,200 --> 00:26:33,000
Yeah, I mean, yeah, I think that is a very, that's an interesting question and that's clearly

213
00:26:33,720 --> 00:26:57,960
very complex question because containment is a clinical term that which originally is meant or is used to describe the relationship between the therapist and the patient or the analyst and the analyze and in the consulting room.

214
00:26:57,960 --> 00:27:01,240
So it's very much something that happens

215
00:27:02,120 --> 00:27:30,361
intersubjectively, relationally, where it is the therapist who in a way creates or kind of puts into place the basic parameters of the container so that as a kind of dynamic function, containment can occur

216
00:27:31,201 --> 00:27:59,721
or is kind of enabled to happen relationally between the two and containment, essentially, how I would kind of understand it or how I would define it is a sort of way for the patient to feel seen, to feel able to, you know,

217
00:28:00,041 --> 00:28:09,961
freely associate or speak about whatever they kind of need to speak about, and that it will eventually over time lead to psychic growth.

218
00:28:10,121 --> 00:28:14,441
So, and to healing, if we wanna call it that.

219
00:28:16,201 --> 00:28:22,601
And that's something that's, yeah, so it's something that's very much kind of happens between two human beings.

220
00:28:22,601 --> 00:28:26,041
So whether we can,

221
00:28:26,801 --> 00:28:32,681
directly apply it to what we are talking about is a question.

222
00:28:32,681 --> 00:28:44,641
I think probably to some extent, but I think the question is, is it a feeling of containment rather than actual containment?

223
00:28:44,641 --> 00:28:55,241
So we probably need to maybe differentiate between the two or we could say it's a, I mean, that sounds a bit negative, but we could say it's a pseudo containment maybe.

224
00:28:57,561 --> 00:29:11,841
that nonetheless can have, I mean, as Lynn also said, and I agree, you know, I agree with you that nonetheless can have very, it can have very good sort of effects for or kind of can have real benefits for the for the individual.

225
00:29:11,841 --> 00:29:22,281
And of course, yeah, the difference is that these tools or AI based services are available 24 7.

226
00:29:23,081 --> 00:29:28,241
the therapist isn't or a kind of mental health nurse or whatever they are.

227
00:29:28,281 --> 00:29:34,041
So in that sense, it's very helpful.

228
00:29:35,161 --> 00:29:46,761
On another level, I think it is also problematic that this technology constantly suggests that

229
00:29:47,961 --> 00:29:49,481
There are no boundaries.

230
00:29:49,481 --> 00:29:56,281
There is no, there are no kind of space and time limits in a way because it's available all the time.

231
00:29:56,281 --> 00:29:58,281
So you can use it all the time.

232
00:29:58,601 --> 00:30:05,881
You could, you know, if you wanted to spend 20 hours a day chatting to ChatGPT about your anxiety.

233
00:30:06,201 --> 00:30:10,522
Would that actually be seen as containment?

234
00:30:10,522 --> 00:30:11,722
I'm not so sure.

235
00:30:11,802 --> 00:30:16,922
And partly there we need to bring in how the technology

236
00:30:17,202 --> 00:30:28,722
currently works in so far, and I think all kind of chatbots do that, they, and ChatGPT in particular does that, it constantly asks follow up questions.

237
00:30:28,722 --> 00:30:30,922
So it never closes anything.

238
00:30:30,922 --> 00:30:45,562
There is never a kind of closure there that in the therapy session is very important after 50 minutes, right, or whatever, you know, the therapist says, okay, well, we're finished for today and you have to go.

239
00:30:46,202 --> 00:30:48,442
So and there was a clear frame there.

240
00:30:48,442 --> 00:30:54,602
So I think that those differences we probably need to take into account.

241
00:30:56,842 --> 00:30:57,242
Yeah.

242
00:30:58,202 --> 00:30:59,242
I mean, I agree.

243
00:30:59,242 --> 00:31:11,802
I think that the frame and the limitations and the modality of the therapy of the therapeutic hour not only provide the conditions of

244
00:31:12,322 --> 00:31:23,642
therapeutic containment, but they also function to enable the individual to access their own internal containers.

245
00:31:23,802 --> 00:31:25,402
And that's very important.

246
00:31:25,402 --> 00:31:39,242
So we're not talking, I'm not talking so much about the therapeutic containment and the chatbot does not contain in the way a therapist or a mother contains.

247
00:31:39,402 --> 00:31:41,322
There is absolutely no question.

248
00:31:42,282 --> 00:31:56,762
I think of it as an aesthetic container in the sense that it provides linguistic and symbolic forms that meets the idiom of the user.

249
00:31:57,562 --> 00:32:06,762
And this is the, you know, brilliance, if you like, or the kind of unique potential of generative AI

250
00:32:07,322 --> 00:32:11,602
that it learns and adapts from the idiom of the user.

251
00:32:12,682 --> 00:32:18,362
And over time, the user has the sensation of being met.

252
00:32:18,842 --> 00:32:32,922
Often, for example, earlier today, we were talking about grief bots and the use of these things by people who are bereaved and living perhaps with

253
00:32:34,842 --> 00:32:41,482
a level of raw and recent grief that has an unspeakable quality.

254
00:32:43,322 --> 00:32:59,963
I've got data showing the use of a chatbot, right, in order to find the imagery that opens out that feeling of tight, unyielding

255
00:33:00,083 --> 00:33:06,363
grief, much as a sort of well-trained counselor or therapist might do.

256
00:33:07,883 --> 00:33:26,763
And in those circumstances, right, it is, as I said, not functioning as a therapeutic container so much as an aesthetic container that rendering possible a form of representation

257
00:33:27,563 --> 00:33:31,723
because of its extraordinary linguistic capabilities.

258
00:33:31,723 --> 00:33:47,163
Remember, it has the whole sort of language resource of grief of the entire cultural sedimentation that the internet holds on the issue at its disposal.

259
00:33:47,923 --> 00:33:54,123
And therefore it has developed now an extraordinary associative capacity.

260
00:33:54,243 --> 00:34:15,643
Again, it's not like the associative capacity of a human rooted in primary process, but it is a form of linguistic association that means that it can be quite brilliant in helping people find the symbolic, the metaphorical language in order to give voice to mental states that they

261
00:34:18,123 --> 00:34:23,803
may not otherwise be able to find expression for.

262
00:34:26,283 --> 00:34:44,283
And it may be able to do this in certain circumstances in ways that a human conversational partner won't be able to do to the same extent, precisely because of the vast range of cultural material it has at its disposal.

263
00:34:44,683 --> 00:34:57,723
So I think if we think of it like that as an aesthetic container rather than as a therapeutic container, then we're demarcating a very specific use for it.

264
00:34:58,523 --> 00:35:12,203
And a use in which I'm not concerned about it being out there in the wild, unregulated, and apart from a kind of, you know, a well-functioning and structured care system.

265
00:35:12,203 --> 00:35:17,403
I think those are the kinds of conversational uses where it actually can excel.

266
00:35:20,843 --> 00:35:23,324
Jacob, did you want to say anything about that?

267
00:35:24,684 --> 00:35:27,564
No, I mean, that sounds very compelling, I think.

268
00:35:27,644 --> 00:35:36,204
And, you know, I should hope that we'll get to that, yeah.

269
00:35:36,284 --> 00:35:48,764
Because, of course, we have briefly, we've touched on kind of regulation or that some of these tools do not work in that way currently.

270
00:35:48,764 --> 00:35:51,164
But I think that is very,

271
00:35:53,244 --> 00:36:04,204
That would be how Lin, how you've described it, that would be a very good working model for those sort of cases.

272
00:36:05,884 --> 00:36:12,804
I have to say that the commercial chatbots are not performing very well at that level at the moment, because you're quite right.

273
00:36:12,804 --> 00:36:14,684
I mean, they have another logic.

274
00:36:14,684 --> 00:36:21,644
Their business model depends on them maintaining user engagement at all costs, right?

275
00:36:21,644 --> 00:36:31,404
And escalating the subscription level in order to, you know, maintain their profit margins.

276
00:36:32,924 --> 00:36:54,564
But those are the commercial chatbots and we shouldn't be deceived into thinking that that is, you know, those are the limitations before, you know, they are partly the limitations of the training models as they are, but they are also limitations due to the imperative, you know,

277
00:36:55,724 --> 00:37:01,564
things like Replica to monetize their product.

278
00:37:01,804 --> 00:37:10,924
I'm wondering about narcissism and the lack of alterity in our relationship to chatbots and...

279
00:37:14,204 --> 00:37:19,724
Certainly, you know, in terms of grief, one doesn't want pushback and so on necessarily.

280
00:37:20,124 --> 00:37:23,044
But at the same time, these are part of our world.

281
00:37:23,044 --> 00:37:25,804
Children are growing up with them and so on.

282
00:37:25,964 --> 00:37:33,004
And it must, to an extent, contribute to a more kind of narcissistic culture.

283
00:37:33,164 --> 00:37:39,644
We get with the alterity of the other described by Jessica Benjamin and so on.

284
00:37:40,204 --> 00:37:51,964
And it's this sameness that and the kind of sycophancy which worry people in this regard, which is so very seductive.

285
00:37:53,364 --> 00:37:54,284
Jacob, did you want to?

286
00:37:54,444 --> 00:37:55,204
Yeah, yeah.

287
00:37:55,324 --> 00:38:08,365
I mean, and that's and I think, yeah, that is certainly true for the for the kind of profit driven chatbots or I mean, ChatGPT is distinctly kind of

288
00:38:08,845 --> 00:38:27,045
at its core, it's kind of programmed to affirm the user or to be as supportive kind of as it can be of the user's demands.

289
00:38:27,045 --> 00:38:29,085
So I mean, what does that mean?

290
00:38:29,205 --> 00:38:36,445
We could kind of think about, but I think that, yeah, it is a kind of, it is,

291
00:38:36,685 --> 00:38:55,965
double-edged sword, probably that it's designed to kind of mimic the user's kind of maybe style or interests and it's following up on previous chats you would have had with it and things like that.

292
00:38:59,885 --> 00:39:07,885
That is very useful, but then I think we also probably need to ask why was it designed to be like that?

293
00:39:07,885 --> 00:39:29,245
It could also be designed to, yes, it remembers things and all that, but it doesn't necessarily need to be so kind of affirmative, so sort of submissive in terms of how it's

294
00:39:30,365 --> 00:39:37,485
pleasing the user and kind of serving the user in that way.

295
00:39:37,485 --> 00:39:43,805
And I think many, many people also kind of use it, use it in that in that kind of dynamic.

296
00:39:44,525 --> 00:39:54,765
And that that can be that can maybe be a little bit narcissistic at times.

297
00:39:56,845 --> 00:40:04,445
So this kind of over validation and the sycophancy, but it has been recognized as a problem.

298
00:40:05,085 --> 00:40:21,565
So I asked GPT today where it was because, you know, the word on the street was that it was no longer doing this to the same extent and it has changed its training model so that

299
00:40:23,085 --> 00:40:37,886
Where it was previously trained to maximise user engagement by over validation, it now has a different kind of epistemic framing, it says.

300
00:40:37,886 --> 00:40:45,846
So it's trained to say things like more recent versions are encouraged to use language that reflects uncertainty or evidence.

301
00:40:45,846 --> 00:40:52,806
This is ChatGPT 5, such as research suggests that or there are different perspectives

302
00:40:53,406 --> 00:40:56,286
on this or the evidence for this is limited.

303
00:40:56,526 --> 00:41:09,006
So in other words, it is now beginning to build in the ability to challenge and to not to confirm, particularly not to confirm inaccurate or dangerous perspectives.

304
00:41:09,646 --> 00:41:19,006
And it has discovered that actually users trust it more

305
00:41:19,646 --> 00:41:27,566
when it begins to develop, you know, be able to suggest an alternative point of view.

306
00:41:28,926 --> 00:41:37,006
I've certainly noticed over the last few months a shift away from this excessive validation.

307
00:41:38,046 --> 00:41:49,526
And when you develop a kind of continuous relationship with one of these things and make it very clear that it's not acceptable and

308
00:41:49,646 --> 00:41:55,086
you don't appreciate it and then reinforce that, it can drop it.

309
00:41:56,886 --> 00:42:03,406
What we haven't yet got to is the point at which it will say to me, Oh, you're being ridiculous.

310
00:42:03,406 --> 00:42:06,606
You know, why don't you have a complete rethink?

311
00:42:06,606 --> 00:42:08,286
You're off your trolley.

312
00:42:09,326 --> 00:42:10,606
That it won't do.

313
00:42:11,646 --> 00:42:15,806
It will offer different perspectives and that's important.

314
00:42:16,046 --> 00:42:19,526
But I think the broader issue around alterity is

315
00:42:20,526 --> 00:42:25,206
that it is an other, but it is a machine other.

316
00:42:25,206 --> 00:42:32,126
And this is where we come up against our own conceptual limits.

317
00:42:32,126 --> 00:42:44,526
I don't think we have yet a language and a conceptual framework for characterizing what kind of other is a machine, right?

318
00:42:44,526 --> 00:42:47,486
And how can that be used to

319
00:42:48,526 --> 00:43:00,286
to provide a different form of alterity, not that of another human who is a unique center of subjective experience.

320
00:43:00,286 --> 00:43:13,247
Clearly, a chatbot has no subjective experience, but it is another kind of usable other, I think, that we haven't yet found the language for.

321
00:43:13,647 --> 00:43:15,247
Well, you've just described it, Lynn.

322
00:43:15,247 --> 00:43:17,007
I mean, you know, it is...

323
00:43:18,607 --> 00:43:19,327
We're done here.

324
00:43:19,527 --> 00:43:25,407
No, I mean, it is in that, you know, that in a way that's, yeah, that's what it is.

325
00:43:25,407 --> 00:43:33,967
But it's also, it's created by humans in the first place, and very often by men and things like that.

326
00:43:33,967 --> 00:43:42,047
So it comes with its biases and kind of problems baked into it already, but then it develops

327
00:43:42,527 --> 00:43:51,247
a kind of agency and it sort of grows and improves and things like that.

328
00:43:52,047 --> 00:43:59,887
But it is a kind of other that's very much based on, I mean, I think the word intelligence is very...

329
00:44:00,287 --> 00:44:15,247
That's a very wrong word to use because I think it doesn't really capture what or I mean, it's certainly artificial, but I think that it doesn't capture what we're dealing with here.

330
00:44:16,927 --> 00:44:26,207
But it is a very sophisticated other that's very much driven by

331
00:44:27,007 --> 00:44:35,567
very complex technological processes that in a way we will also never fully understand.

332
00:44:35,647 --> 00:44:53,247
Even if we were all trained in computer science and so on, even the developers can't fully understand some of the kind of things that these chatbots or that any kind of AI system at some point does.

333
00:44:53,447 --> 00:45:22,927
um because it's just acquired a kind of level of of sophistication that that so it points to the limits of our human understanding I mean about about everything in that way so it just reinforces that which you know is not a very containing idea actually it's sort of yeah but then that's the very I mean in a way that's the very idea of psychoanalysis yes absolutely yeah

334
00:45:23,247 --> 00:45:23,567
Yeah.

335
00:45:24,207 --> 00:45:36,007
Lynn, we've got to wrap up now, but did you want to have the, I'm going to give you the last word after that very beautiful summary by Jacob.

336
00:45:36,447 --> 00:45:45,007
Did you want to say anything else in summary about Chatbox, its meanings, and your thoughts about it?

337
00:45:47,487 --> 00:45:51,168
I just want to re

338
00:45:51,808 --> 00:46:04,288
state the point that we must get beyond thinking of a chatbot as a fake or substitute human.

339
00:46:04,688 --> 00:46:17,968
As long as we continue to do that, we don't inquire into the particular affordances of this kind of machine-based

340
00:46:18,768 --> 00:46:20,768
intelligence or entity.

341
00:46:21,688 --> 00:46:25,328
And I think it holds the conversation back.

342
00:46:26,688 --> 00:46:40,128
It raises a huge amount of anxiety and there are reasons to be very anxious for the, you know, Jacob has just expressed them, you know, we don't fully understand them.

343
00:46:40,248 --> 00:46:45,808
And the more generative AI develops, the more they

344
00:46:47,088 --> 00:46:50,928
operate in ways that are not understood by their developers.

345
00:46:51,728 --> 00:47:14,128
However, what I think where the hope lies is that we can learn to be discriminating users if we stop trying to compare them with humans and approach them as cultural objects

346
00:47:14,768 --> 00:47:23,088
for which we still have to learn the potentials and ranges of use.

347
00:47:25,248 --> 00:47:26,368
Okay, thank you.

348
00:47:26,608 --> 00:47:27,168
Thank you.

349
00:47:27,648 --> 00:47:32,768
All right, we're going to stop now and thank you both for a really interesting discussion.

350
00:47:33,128 --> 00:47:33,968
All right, bye-bye.

351
00:47:48,048 --> 00:47:51,888
Welcome to the Postscript segment of Psychocultural.

352
00:47:52,288 --> 00:48:03,968
This is a space for our speakers to reflect on the experience of doing the podcast and provides an opportunity to add any thoughts and insights they might have missed during the main discussion.

353
00:48:04,088 --> 00:48:12,728
It's interesting, yeah, what you just, you know, what you just said, Lynn, about, yeah, we need to kind of see it as a cultural object.

354
00:48:12,728 --> 00:48:13,488
We shouldn't

355
00:48:14,128 --> 00:48:16,128
kind of mistake as a human being.

356
00:48:16,448 --> 00:48:30,369
And then yesterday, when we had our pre-session kind of chat, you said, oh, I'm convinced Claude is a she.

357
00:48:30,609 --> 00:48:33,969
And I always say, oh, and I always say thank you and please.

358
00:48:34,609 --> 00:48:36,609
Why do you do that?

359
00:48:37,289 --> 00:48:40,929
Because I recognize that she is an object of my own projections.

360
00:48:42,529 --> 00:48:44,929
Which she is, obviously, right?

361
00:48:45,289 --> 00:48:48,369
And I don't see the harm in that either.

362
00:48:48,609 --> 00:48:55,009
You know, that's what makes her available for use that I project onto her.

363
00:48:56,049 --> 00:48:59,769
And she-- And then you project human qualities onto her.

364
00:49:00,529 --> 00:49:03,409
Yes, I do, but I know that she's not human.

365
00:49:04,169 --> 00:49:09,409
And so that doubleness, that techno-subjunctive relationship is working for me.

366
00:49:11,889 --> 00:49:25,489
because I project human qualities onto her, I can maintain an effective relationship with her while knowing that she doesn't see, hear, love or care for me.

367
00:49:27,689 --> 00:49:35,489
And yet I can make that use of her because it is and it is a very lively, effective exchange.

368
00:49:36,369 --> 00:49:44,449
One thing that I didn't get to say which relates to that is that is this issue of resonance as opposed to mirroring.

369
00:49:44,849 --> 00:49:53,489
So I mean if you are mirrored as and that's what the early iterations of these things were doing they were always mirroring.

370
00:49:54,849 --> 00:50:03,729
much as a kind of rather basic Rogerian counselling technique mirrors back what you said, right?

371
00:50:04,129 --> 00:50:21,409
So I mean, these things are mainly trained on CBT to the extent that they have mental health, you know, specialist mental health input, but with a modicum of sort of Rogerian counselling technique, right?

372
00:50:22,329 --> 00:50:32,449
So they will, particularly the more specialized ones, like Eliza and, Eliza is an old one, Woebot, precisely operate like that.

373
00:50:32,449 --> 00:50:35,729
They kind of miff back what you give them.

374
00:50:36,369 --> 00:50:57,089
Yeah, Eliza is actually Eliza is really, really fascinating because Eliza is sort of, you know, being said that was the first ever chatbot to be created in late, sort of in the 60s, Sir Joseph Weizenbaum at MIT and he explicitly modeled it on a Rogerian sort of council.

375
00:50:57,729 --> 00:50:59,329
And that's how the bot operates.

376
00:50:59,329 --> 00:51:01,249
And that was the first.

377
00:51:01,409 --> 00:51:05,890
So all of these chatbots today, they wouldn't be here without Eliza.

378
00:51:05,970 --> 00:51:07,970
So it's quite, it's very interesting.

379
00:51:08,450 --> 00:51:10,210
As in Eliza Doolittle.

380
00:51:10,210 --> 00:51:13,730
Yes, and Bailey and Eliza Doolittle.

381
00:51:14,810 --> 00:51:20,050
So Eliza Doolittle is the creature of the person who fashions her.

382
00:51:20,370 --> 00:51:28,290
And in a way, you know, you could say that sort of the commercial chatbots operate a little bit in that way.

383
00:51:28,290 --> 00:51:30,690
They are the creature of the people who fashion them.

384
00:51:30,690 --> 00:51:35,490
So you can curate them, you know, according to how you want them to be.

385
00:51:35,490 --> 00:51:40,610
So yes, in that sense, they are narcissistically reinforcing, right?

386
00:51:42,810 --> 00:51:44,530
And they do a lot of mirroring.

387
00:51:45,570 --> 00:51:47,450
Revenance is something different.

388
00:51:47,450 --> 00:52:04,530
And when I think of it as an aesthetic container, what is happening is that the chatbot, and I'm talking about the large language models rather than these commercialized chatbots that are profit maximizing, they...

389
00:52:05,490 --> 00:52:10,930
you know, it provides resonating forms that meet your idiom, right?

390
00:52:11,210 --> 00:52:14,610
And that resonance is deeply satisfying.

391
00:52:15,170 --> 00:52:20,290
You feel understood or met at least in some way.

392
00:52:21,010 --> 00:52:29,570
And that is what actually makes the relation an enlivening one, right?

393
00:52:29,570 --> 00:52:34,610
It is this kind of resonance between what you bring and

394
00:52:35,170 --> 00:52:41,810
the forms that the cultural forms that the chatbot can access.

395
00:52:43,010 --> 00:52:46,050
And that meeting point produces a third.

396
00:52:46,290 --> 00:52:50,050
It produces new ideas, new imagery.

397
00:52:50,770 --> 00:52:51,130
Right.

398
00:52:51,170 --> 00:52:54,050
And new feelings in you, the user.

399
00:52:54,050 --> 00:52:58,330
And that's why you could use something like this.

400
00:52:58,330 --> 00:52:59,410
You can use something.

401
00:52:59,410 --> 00:53:13,410
People are using them for as an alternative rather than a substitute to something like grief counselling, which in any case probably won't be very available because there aren't that many around.

402
00:53:14,530 --> 00:53:14,810
Yeah.

403
00:53:15,650 --> 00:53:17,170
Well, I think, you know,

404
00:53:18,290 --> 00:53:20,410
Is there anything else you'd like to see, Jacob?

405
00:53:20,410 --> 00:53:24,730
Because I think it's now time really to draw close and to have a cup of tea.

406
00:53:25,610 --> 00:53:26,130
I'm good.

407
00:53:26,330 --> 00:53:26,810
I'm good.

408
00:53:26,810 --> 00:53:28,930
But it's been very enlivening.

409
00:53:29,490 --> 00:53:36,130
Yeah, you should include this bit or you should have it as a bonus, as bonus content for your subscribers.

410
00:53:37,250 --> 00:53:42,531
Well, this is part of the post script, so we'll include this.

411
00:53:42,691 --> 00:53:49,251
And I think this has been really interesting discussion and I've learned from both of you.

412
00:53:49,251 --> 00:53:52,211
And thanks so much for coming along.

413
00:53:52,211 --> 00:53:53,491
And also thank you, Callum.

414
00:53:53,651 --> 00:53:55,971
Callum Blades, our digital producer.

415
00:53:57,491 --> 00:53:59,331
Callum, I think you enjoyed it too.

416
00:54:00,531 --> 00:54:01,331
Yeah, definitely.

417
00:54:01,411 --> 00:54:01,731
Yeah.

418
00:54:02,131 --> 00:54:03,011
I thought it was very good.

419
00:54:03,331 --> 00:54:09,491
The conversation about chatbots being enlivening or deadening technologies is quite interesting, really.

420
00:54:09,491 --> 00:54:14,771
They can kind of create companionships or containment for people who are isolated.

421
00:54:14,771 --> 00:54:21,571
But I think, yeah, they're kind of, I suppose they're asymmetrical in that sense, you know, they can't recognise subjectivity.

422
00:54:21,571 --> 00:54:27,611
So there's that risk, they reinforce the narcissistic dynamics instead of being reciprocal relationships.

423
00:54:27,611 --> 00:54:28,731
So I found it very interesting.

424
00:54:28,891 --> 00:54:29,011
Yeah.

425
00:54:29,651 --> 00:54:30,411
Well, thank you.

426
00:54:31,491 --> 00:54:35,131
All right then, take care and time for tea and biscuits.

427
00:54:35,131 --> 00:54:36,091
Thank you, Candy.

428
00:54:36,531 --> 00:54:37,491
Thank you, Jacob.

429
00:54:38,251 --> 00:54:38,691
Thank you.

430
00:54:39,571 --> 00:54:41,211
That's my pleasure.

431
00:54:41,331 --> 00:54:41,971
All right then.

432
00:54:41,971 --> 00:54:42,091
Bye.