Popular in Psychology

Most Cited

The adapted mind

Dalhousie University Jerome Barkow

Psychology

Consumers’ pursuit of material and experiential purchases

Cornell University Tom Gilovich

Experimental Psychology

Time to investigate human visual shape perception

University of Montreal Martin Arguin

Most Cited

James S. Jackson and the program for research on Black Americans

Harvard University David R. Williams

Psychology

Violence and Mental Illness: Rethinking Risk Factors and Enhancing Public Safety

Duke University Eric Elbogen

Cognitive Psychology

Cognitive remediation – where are we now and what should we do next?

King's College London Til Wykes

Cognitive Psychology

Harnessing Plasticity in the Atypical Developing Brain using Neurostimulation

University of Surrey Roi Cohen Kadosh

A quasi-comprehensive exploration of the mechanisms of spatial working memory

View more insights in Social Sciences - Psychology - Cognitive Psychology

View more figures from The Chinese University of Hong Kong

View the page of Liqiang Huang

Read the Study

Next in Cognitive Psychology

How the Social Becomes the Biological: The Interaction Between the Genome and the Environment

Gillian Einstein | University of Toronto

Paradoxes in a prism: Reflections on the omnipotent passivity and omniscient oblivion of schizophrenia

Clara Humpston | University of York

Why do some spatial patterns stick in our memory more than others? Liqiang Huang delves into the intricate mechanisms behind spatial working memory by integrating multiple theories into a single conceptual model.

Image courtesy of interviewee. October 4, 2024

Log-in or Sign-up to Faculti
Currently viewing this subject insight as a guest. You have insight(s) remaining for this month. Login to view 8000+ figures on the platform.

Generating Workspace...

Export as PDF

This video discusses the concept of spatial working memory, which refers to our ability to remember spatial patterns and visual information at a given moment. It highlights that working memory can be divided into two components: one for visual information (like colors and locations) and another for verbal information. The speaker is studying cognitive factors influencing how well we can remember different objects and patterns, emphasizing the importance of regularity and organization.

The speaker aims to create an overarching framework to integrate various mechanisms discovered in past studies. They express frustration over the fragmented nature of psychology research, where new findings often overshadow older ones without forming a cohesive understanding. To address this, the speaker built an online platform to conduct large-scale experiments disguised as games, allowing data collection from a broader audience.

During the COVID pandemic, a collaboration with a game company helped enhance data collection significantly. The experiments involve presenting participants with visual memory tasks, where they must recall the locations of dots after a brief presentation.

The speaker believes that traditional methods in psychology could benefit from insights drawn from artificial intelligence, particularly using neural networks to better model and explain working memory processes. They discuss ongoing collaborative research into understanding emotions and visual perception, aiming to improve the breadth and applicability of psychological studies.

The overall goal is to produce more relevant, real-world applications based on comprehensive, high-quality data, addressing key questions in psychology that remain unresolved due to a lack of adequate research frameworks.

Language:

Export as PDF

Export as PDF
Here are five key points from the video regarding spatial working memory and its study in psychology:

1. **Definition and Components of Working Memory**:
Spatial working memory is a form of working memory that focuses on visual and spatial information, such as remembering locations and patterns. Working memory generally has two parts: one for visual information and another for verbal information. Spatial working memory specifically relates to the visual aspect.

2. **Mechanisms of Memory and Cognitive Factors**:
The study of spatial working memory encompasses various mechanisms of remembering, such as chunking and categorization. The video highlights the fragmented nature of psychological research, which often focuses on isolated factors instead of an integrated framework for understanding memory.

3. **Challenges in Psychological Research**:
There is a lack of consensus in the psychological community regarding a comprehensive framework for memory. Despite numerous studies, the author notes the difficulty of synthesizing findings into a coherent model to guide practical applications, such as teaching techniques for better memory retention.

4. **Innovative Data Collection via Gaming**:
The author describes an innovative approach to data collection through a partnership with a game company, allowing for the collection of significant amounts of data during the COVID-19 pandemic. This method attracted many participants and extensive data for analyzing spatial working memory tasks.

5. **Combining Cognitive Models with Neural Networks**:
The ongoing research aims to develop a more complex and holistic cognitive model by integrating insights from artificial intelligence, particularly neural networks. The author believes that modern psychology can benefit from a better understanding of various cognitive processes to create an overarching framework, moving beyond the fragmented nature of current studies.

Export as PDF
Here are some key quotes and concepts extracted from the video:

1. **Definition of Spatial Working Memory**:
- "Spatial working memory is basically what we remember at this moment, like a spatial pattern or something in front of you."

2. **Components of Working Memory**:
- "Working memory is normally assumed to include two parts: one part for visual information... and another part for verbal information."

3. **Nature of Cognitive Studies**:
- "In psychology studies, unlike physics, things tend to be fragmented."

4. **Framework Challenge**:
- "Unfortunately, there is no overall framework or at least there's no overall framework that everyone agrees with."

5. **Research Approach**:
- "The particular study I'm trying to... is to try to build up an overall framework."

6. **Experimental Psychology Standards**:
- "In experimental psychology, we normally only need a few, or like, tens of participants and each participating for one hour."

7. **Data Collection Strategy**:
- "I spent some time to try to build up an online platform to package those experiments as games and attract people to play those games."

8. **Collaboration with Gaming Industry**:
- "...some guy approached me... I work for this game company... maybe they want to make a friend and then they offer to implement my experiment as one component in their game."

9. **Neural Networks**:
- "We want to use this neural network...to combine different mechanisms... we build a neural network to do that and then train the neural network to predict people's working memory on different patterns."

10. **Psychology's Fragmentation Issues**:
- "I have this idea or belief now that there is something wrong... with the way how psychology has been studying things because mainly because it's fragmented."

11. **Aim for Comprehensive Models**:
- "We never evolve into an overall framework... maybe we can borrow this insight from Artificial Intelligence and try to build something that's more complex and more complete."

12. **Exploration of Emotions**:
- "In psychology, there is this model on emotion... we tend to believe that neither of those are complete."

These quotes highlight the speaker's exploration of spatial working memory, the need for a unified framework in psychology, the importance of large-scale studies, and the integration of technology like neural networks into psychological research.

X-TIMESTAMP-MAP=LOCAL:00:00:00.000,MPEGTS:90000

1
00:00:00.000 --> 00:00:03.245
Spatial working memory is basically what we

2
00:00:03.845 --> 00:00:05.525
remember at at this moment.

3
00:00:06.425 --> 00:00:10.965
Um, like a spatial pattern or something in front of you.

4
00:00:11.505 --> 00:00:15.045
Um, in psychology there is, working memory,

5
00:00:15.105 --> 00:00:16.125
the basic concept,

6
00:00:16.985 --> 00:00:21.725
but, working memory, is normally assumed to,

7
00:00:21.795 --> 00:00:22.925
include two parts.

8
00:00:23.665 --> 00:00:27.245
One part for visual information, like you remember,

9
00:00:27.685 --> 00:00:30.645
remember colors or remember a lot of loca.

10
00:00:30.745 --> 00:00:34.925
Um, a few locations, like if I like po, point a few

11
00:00:35.525 --> 00:00:36.965
locations and you can remember that.

12
00:00:37.665 --> 00:00:40.685
And another part is for verbal information.

13
00:00:40.835 --> 00:00:45.125
Like I, speak something a KCY,

14
00:00:45.125 --> 00:00:46.325
you can remember those letters.

15
00:00:47.915 --> 00:00:51.175
And, that's the two parts of working memory.

16
00:00:51.275 --> 00:00:54.255
And the, special working memory is one part of the

17
00:00:54.995 --> 00:00:56.615
visual, part of working memory.

18
00:00:56.685 --> 00:01:00.655
Like the, what we remember about the special information,

19
00:01:00.765 --> 00:01:03.935
what I'm studying is, cognitive factors,

20
00:01:04.845 --> 00:01:06.895
basically the stimuli side.

21
00:01:07.035 --> 00:01:11.015
So what, kind of objects can we remember better?

22
00:01:11.195 --> 00:01:13.775
And, in what way do we remember them?

23
00:01:14.725 --> 00:01:17.385
So for example,

24
00:01:18.085 --> 00:01:20.945
if I present a few things in front of you

25
00:01:21.485 --> 00:01:22.785
and they're very regular,

26
00:01:23.085 --> 00:01:25.945
you can remember those more easily than a

27
00:01:25.945 --> 00:01:27.145
few scattered thing.

28
00:01:27.725 --> 00:01:32.025
In. In this study, I'm trying to combine things like,

29
00:01:32.315 --> 00:01:35.745
there are about, 10 to 15 different mechanisms

30
00:01:35.745 --> 00:01:39.745
that has been dec discovered in previous studies.

31
00:01:40.565 --> 00:01:43.625
Uh, like what I just mentioned about regularity.

32
00:01:43.625 --> 00:01:45.265
That's one thing, like chunking

33
00:01:45.605 --> 00:01:48.065
and of, of course, the other factors, like, like,

34
00:01:48.735 --> 00:01:52.585
categorical information or like base integration.

35
00:01:52.685 --> 00:01:55.105
Uh, there are all those fancy terms,

36
00:01:55.365 --> 00:01:58.485
but, for our purpose,

37
00:01:59.265 --> 00:02:00.365
the important thing

38
00:02:00.395 --> 00:02:02.245
that they're all different from each other.

39
00:02:02.905 --> 00:02:06.565
And, so, as you may

40
00:02:06.585 --> 00:02:09.445
or may not be aware, in the psychology studies,

41
00:02:09.745 --> 00:02:13.485
unlike physics, things tends to be fragmented.

42
00:02:13.555 --> 00:02:16.525
Like people discuss, discover one thing

43
00:02:16.585 --> 00:02:18.125
and they discover another thing.

44
00:02:18.545 --> 00:02:23.205
So it's more always, in those topics related to

45
00:02:23.955 --> 00:02:26.205
like, other things,

46
00:02:26.205 --> 00:02:28.125
but it also applies to working memory.

47
00:02:28.745 --> 00:02:32.765
For example, my wife, always ask me like,

48
00:02:32.785 --> 00:02:36.805
you study memory, so how do we like teach our child

49
00:02:37.025 --> 00:02:38.805
so they can remember things better?

50
00:02:38.865 --> 00:02:42.925
And I told her they all kind of different studies,

51
00:02:43.905 --> 00:02:44.925
and this and that,

52
00:02:45.305 --> 00:02:49.605
but unfortunately there is no overall like a framework

53
00:02:49.705 --> 00:02:52.165
or at least there's no overall framework

54
00:02:52.165 --> 00:02:53.565
that everyone agree with.

55
00:02:54.425 --> 00:02:58.805
So the ca the particular study I'm trying to,

56
00:02:59.165 --> 00:03:01.685
I have done recently, and I'm trying to do,

57
00:03:01.755 --> 00:03:05.645
more things on this, is to try to build up,

58
00:03:06.275 --> 00:03:07.325
overall framework.

59
00:03:08.145 --> 00:03:12.445
So to combine things together and experimental psychology.

60
00:03:13.145 --> 00:03:17.485
Uh, the typical standard is we run,

61
00:03:18.035 --> 00:03:21.205
participant for something from 10 to 20,

62
00:03:21.475 --> 00:03:23.725
sometimes maybe 40, 50.

63
00:03:24.345 --> 00:03:28.205
But, normally we don't need to go more than like,

64
00:03:28.675 --> 00:03:32.485
like a hundred to establish a point,

65
00:03:32.585 --> 00:03:37.205
unlike some other area, like, they need larger study.

66
00:03:37.305 --> 00:03:39.405
But in experimental, psychology,

67
00:03:39.465 --> 00:03:42.285
we normally only need a few,

68
00:03:42.345 --> 00:03:44.405
or like, tens of participants

69
00:03:44.865 --> 00:03:47.085
and each participating for one hour.

70
00:03:47.865 --> 00:03:51.965
And, so as you know, like, when we want to

71
00:03:52.725 --> 00:03:57.165
explicit a complex pattern, we need a lot

72
00:03:57.165 --> 00:03:58.845
of statistical power, a lot

73
00:03:58.845 --> 00:04:01.165
of information to figure things out.

74
00:04:01.165 --> 00:04:03.485
That's why we need a large scale experiment.

75
00:04:04.065 --> 00:04:07.385
So I spent, some time to try

76
00:04:07.385 --> 00:04:12.105
to build up a online platform to like,

77
00:04:12.465 --> 00:04:15.385
packages those, experiment as games

78
00:04:15.645 --> 00:04:17.545
and attract people to play those games.

79
00:04:18.405 --> 00:04:23.345
And, interesting, thing is that this doesn't work

80
00:04:23.895 --> 00:04:24.985
very well at first

81
00:04:24.985 --> 00:04:27.305
because, it takes a long time to attract people.

82
00:04:27.845 --> 00:04:31.385
And, but at some point, some guy approached me

83
00:04:31.445 --> 00:04:35.065
and, say, what are you exactly trying

84
00:04:35.065 --> 00:04:36.185
to do with this?

85
00:04:36.885 --> 00:04:37.945
Uh, those

86
00:04:38.285 --> 00:04:39.625
and then,

87
00:04:40.125 --> 00:04:43.065
really look like real games not so interesting.

88
00:04:43.605 --> 00:04:44.865
And, but,

89
00:04:44.885 --> 00:04:46.905
and obviously you're not gonna make money out of it.

90
00:04:47.325 --> 00:04:49.905
And why, why are you doing this? And I just told him.

91
00:04:50.525 --> 00:04:52.345
And then he told me, okay, I work

92
00:04:52.365 --> 00:04:54.945
for this game company, real game company.

93
00:04:55.645 --> 00:04:58.585
So maybe like, as a,

94
00:04:58.735 --> 00:05:03.025
because many of the, zen are also interested,

95
00:05:03.365 --> 00:05:04.425
in psychology.

96
00:05:04.425 --> 00:05:05.745
So maybe they want to make a friend

97
00:05:05.765 --> 00:05:07.745
and then they offer to implement

98
00:05:08.565 --> 00:05:11.025
my experiment are one component in their game,

99
00:05:11.665 --> 00:05:13.985
actually popular game, like,

100
00:05:15.095 --> 00:05:18.195
and then they did it, it actually cost me nothing

101
00:05:18.415 --> 00:05:23.355
to collect data for like, let me see how to,

102
00:05:23.355 --> 00:05:25.155
hundred of thousand of hours.

103
00:05:26.425 --> 00:05:31.085
And, like, of course not regular hours

104
00:05:31.085 --> 00:05:33.925
because some just play for a few like a minute

105
00:05:34.065 --> 00:05:36.205
and then left some play for longer.

106
00:05:36.705 --> 00:05:41.205
But, all come together like, hundreds of thousands

107
00:05:41.205 --> 00:05:42.565
of like hours.

108
00:05:42.745 --> 00:05:46.565
And then I got that data and then analyze that.

109
00:05:46.945 --> 00:05:48.685
And so do this task.

110
00:05:49.285 --> 00:05:52.885
I present, a few dots in, like a

111
00:05:53.505 --> 00:05:54.645
six by six array.

112
00:05:55.265 --> 00:05:56.845
And then, they

113
00:05:56.885 --> 00:05:59.405
Disappear within a second or so.

114
00:05:59.505 --> 00:06:01.045
And then, like,

115
00:06:01.815 --> 00:06:05.645
after another second, the participant have to point out

116
00:06:05.695 --> 00:06:08.285
where are those locations they have seen,

117
00:06:08.355 --> 00:06:10.205
it's a standard working memory task.

118
00:06:10.665 --> 00:06:14.165
So because, by the time we did that,

119
00:06:14.465 --> 00:06:17.125
the game company helped me to do that,

120
00:06:17.345 --> 00:06:18.925
it was in the covid time.

121
00:06:19.545 --> 00:06:23.885
So they, they, like, they have this game related to covid,

122
00:06:24.145 --> 00:06:26.605
so they, okay, so maybe they package that

123
00:06:26.605 --> 00:06:29.205
as a one little component in their game.

124
00:06:29.625 --> 00:06:33.165
And, like you are catching those wires, like,

125
00:06:33.305 --> 00:06:34.765
something like that.

126
00:06:35.065 --> 00:06:37.845
The reason why we want to use this neural network is

127
00:06:37.845 --> 00:06:42.165
because when we want to combine different mechanisms,

128
00:06:42.785 --> 00:06:45.125
we don't know what we are expecting

129
00:06:45.235 --> 00:06:48.485
because that's a different perspective from the traditional

130
00:06:48.925 --> 00:06:53.205
experimental psychology in which you like, just,

131
00:06:53.205 --> 00:06:54.285
test a hypothesis

132
00:06:54.305 --> 00:06:56.965
and you say, okay, hypothesis is significant,

133
00:06:57.705 --> 00:06:59.365
and then you establish a point.

134
00:06:59.825 --> 00:07:02.645
But now what I'm trying to do, I want to build a model

135
00:07:02.905 --> 00:07:05.645
to explain the data as much as possible.

136
00:07:06.025 --> 00:07:08.965
So there's no always benchmark on

137
00:07:08.985 --> 00:07:11.845
how good is good enough.

138
00:07:12.075 --> 00:07:14.845
Obviously you will not be able to explain everything.

139
00:07:15.745 --> 00:07:19.605
Um, and, you'll always be able to explain something.

140
00:07:19.705 --> 00:07:22.405
So what, how good is good is,

141
00:07:22.495 --> 00:07:24.605
quite weak, quite ambiguous.

142
00:07:25.185 --> 00:07:27.725
So I think, okay, maybe,

143
00:07:28.525 --> 00:07:32.235
nowadays neural network is like,

144
00:07:33.155 --> 00:07:36.035
a good standard, good standard for like,

145
00:07:36.375 --> 00:07:37.435
almost everything.

146
00:07:37.815 --> 00:07:40.795
And then maybe we build a neural network to do that

147
00:07:41.255 --> 00:07:46.155
and then train neur neural network to predict like, the,

148
00:07:47.225 --> 00:07:50.355
like, people's working memory on different patterns

149
00:07:50.655 --> 00:07:52.075
and see how much can the,

150
00:07:52.255 --> 00:07:54.315
can the neural network explain the data?

151
00:07:55.015 --> 00:07:58.075
And then we compare our cognitive model.

152
00:07:58.225 --> 00:08:00.915
It's more complex than the typical one,

153
00:08:00.915 --> 00:08:03.835
but it's still a cognitive model in and fully interpretable.

154
00:08:04.335 --> 00:08:06.835
Uh, and then compare that to neural network.

155
00:08:07.215 --> 00:08:10.635
Uh, it's, didn't do as well as neural network,

156
00:08:10.815 --> 00:08:11.835
but it's get close.

157
00:08:12.055 --> 00:08:15.395
So that's like, so we think, okay,

158
00:08:15.395 --> 00:08:17.155
that's good enough because it's already

159
00:08:17.445 --> 00:08:18.635
close to a neural network.

160
00:08:19.075 --> 00:08:21.195
I have this, idea

161
00:08:21.375 --> 00:08:25.355
or belief now that, there is something wrong

162
00:08:25.575 --> 00:08:29.235
or at least incomplete with the way how,

163
00:08:29.885 --> 00:08:32.515
psychology has been studying things because mainly

164
00:08:32.515 --> 00:08:35.115
because it's fragmented and,

165
00:08:36.315 --> 00:08:38.095
and we're not going anywhere, like

166
00:08:38.095 --> 00:08:40.215
after 20 years we discover a lot of things

167
00:08:40.275 --> 00:08:43.175
and then people forget about those older things

168
00:08:43.355 --> 00:08:45.255
and then keep digging new things.

169
00:08:45.395 --> 00:08:48.335
We never evolve into overall framework.

170
00:08:49.035 --> 00:08:51.615
So that's why I think, okay, maybe we can borrow

171
00:08:52.285 --> 00:08:54.255
this insight from the

172
00:08:54.965 --> 00:08:57.245
Artificial intelligence and try to build something

173
00:08:57.245 --> 00:09:00.605
that's more complex and, but more complete.

174
00:09:01.345 --> 00:09:04.925
So for that purpose, that's why we need,

175
00:09:05.495 --> 00:09:06.685
large scale experiment.

176
00:09:06.685 --> 00:09:08.485
We need a lot of data for that purpose.

177
00:09:09.185 --> 00:09:11.845
So with this general idea, I started

178
00:09:11.915 --> 00:09:14.525
with the topic I'm more familiar with,

179
00:09:14.705 --> 00:09:18.925
but now with that in hand now, with a bunch

180
00:09:18.985 --> 00:09:21.645
of collaborators, we are moving towards something

181
00:09:21.645 --> 00:09:26.245
that's more real world relevant, like also like,

182
00:09:26.255 --> 00:09:27.805
using large scale experiment.

183
00:09:28.345 --> 00:09:32.725
Um, for example, in this one study, we,

184
00:09:33.325 --> 00:09:36.325
a, a few collaborators, collaborator, collaborators

185
00:09:36.385 --> 00:09:39.365
and I, we are trying to figure out what is,

186
00:09:40.505 --> 00:09:44.605
how we actually, represent emotion in our,

187
00:09:44.865 --> 00:09:46.325
in our ma in our mind.

188
00:09:47.065 --> 00:09:48.525
Uh, you may have heard that,

189
00:09:48.665 --> 00:09:52.285
in psychology there is this model, on emotion.

190
00:09:52.515 --> 00:09:56.045
Some of them is divided into different categories like

191
00:09:56.045 --> 00:09:58.325
happiness or sadness, et cetera.

192
00:09:58.795 --> 00:10:01.125
Some of them like,

193
00:10:02.065 --> 00:10:05.325
map all those different emotions into in A 2D space.

194
00:10:05.865 --> 00:10:08.565
But, we tend to believe that neither

195
00:10:08.565 --> 00:10:09.885
of those are complete.

196
00:10:10.345 --> 00:10:14.405
And the reason why we don't have anything better,

197
00:10:14.465 --> 00:10:17.205
for this very important question is we don't have a large

198
00:10:17.295 --> 00:10:22.045
scale, high quality data for us to build a model on it.

199
00:10:22.785 --> 00:10:25.125
And, that's what we are doing now

200
00:10:25.985 --> 00:10:29.485
and the few other things on visual beauty

201
00:10:29.905 --> 00:10:31.965
and also on visual expression.

202
00:10:32.105 --> 00:10:36.165
And that's what, we like, I I just mentioned.

203
00:10:36.185 --> 00:10:39.605
And also on like, language understanding

204
00:10:39.865 --> 00:10:41.285
or the different things.

205
00:10:41.825 --> 00:10:46.405
So hopefully those will make, like, larger impact

206
00:10:46.545 --> 00:10:49.085
and hopefully we'll get into some real

207
00:10:49.085 --> 00:10:50.125
world applications.

208
00:10:50.905 --> 00:10:52.125
Uh, but we are working on that.

Understanding Spatial Working Memory: A Dual-Faceted Approach

00:00:00 - Spatial working memory is basically what we remember at at this moment. like a spatial pattern or something in front of you. in psychology there is, working memory, the basic concept, but, working memory, is normally assumed to, include two parts. One part for visual information, like you remember, remember colors or remember a lot of loca. a few locations, like if I like po, point a few

00:00:37 - And another part is for verbal information. Like I, speak something a KCY, you can remember those letters. And, that's the two parts of working memory. And the, special working memory is one part of the visual, part of working memory. Like the, what we remember about the special information, what I'm studying is, cognitive factors, basically the stimuli side.

00:01:11 - And, in what way do we remember them? So for example, if I present a few things in front of you and they're very regular, you can remember those more easily than a few scattered thing. In. In this study, I'm trying to combine things like, there are about, 10 to 15 different mechanisms that has been dec discovered in previous studies.

Fragmentation in Psychological Research and the Quest for a Unified Framework

00:01:43 - That's one thing, like chunking and of, of course, the other factors, like, like, categorical information or like base integration. there are all those fancy terms, but, for our purpose, the important thing that they're all different from each other. And, so, as you may or may not be aware, in the psychology studies, unlike physics, things tends to be fragmented.

00:02:16 - and they discover another thing. So it's more always, in those topics related to like, other things, but it also applies to working memory. For example, my wife, always ask me like, you study memory, so how do we like teach our child so they can remember things better? And I told her they all kind of different studies,

00:02:45 - but unfortunately there is no overall like a framework or at least there's no overall framework that everyone agree with. So the ca the particular study I'm trying to, I have done recently, and I'm trying to do, more things on this, is to try to build up, overall framework. So to combine things together and experimental psychology. the typical standard is we run,

Challenges in Conducting Large-Scale Experiments in Psychology

00:03:21 - sometimes maybe 40, 50. But, normally we don't need to go more than like, like a hundred to establish a point, unlike some other area, like, they need larger study. But in experimental, psychology, we normally only need a few, or like, tens of participants and each participating for one hour. And, so as you know, like, when we want to

00:03:57 - of statistical power, a lot of information to figure things out. That's why we need a large scale experiment. So I spent, some time to try to build up a online platform to like, packages those, experiment as games and attract people to play those games. And, interesting, thing is that this doesn't work

00:04:24 - because, it takes a long time to attract people. And, but at some point, some guy approached me and, say, what are you exactly trying to do with this? those and then, really look like real games not so interesting. And, but, and obviously you're not gonna make money out of it. And why, why are you doing this? And I just told him. And then he told me, okay, I work for this game company, real game company.

Leveraging Gaming for Memory Research During COVID

00:04:58 - because many of the, zen are also interested, in psychology. So maybe they want to make a friend and then they offer to implement my experiment are one component in their game, actually popular game, like, and then they did it, it actually cost me nothing to collect data for like, let me see how to, hundred of thousand of hours.

00:05:31 - because some just play for a few like a minute and then left some play for longer. But, all come together like, hundreds of thousands of like hours. And then I got that data and then analyze that. And so do this task. I present, a few dots in, like a six by six array. And then, they

00:05:59 - And then, like, after another second, the participant have to point out where are those locations they have seen, it's a standard working memory task. So because, by the time we did that, the game company helped me to do that, it was in the covid time. So they, they, like, they have this game related to covid, so they, okay, so maybe they package that as a one little component in their game.

Integrating Neural Networks with Working Memory Research

00:06:33 - something like that. The reason why we want to use this neural network is because when we want to combine different mechanisms, we don't know what we are expecting because that's a different perspective from the traditional experimental psychology in which you like, just, test a hypothesis and you say, okay, hypothesis is significant, and then you establish a point.

00:07:02 - to explain the data as much as possible. So there's no always benchmark on how good is good enough. Obviously you will not be able to explain everything. and, you'll always be able to explain something. So what, how good is good is, quite weak, quite ambiguous. So I think, okay, maybe, nowadays neural network is like,

00:07:36 - almost everything. And then maybe we build a neural network to do that and then train neur neural network to predict like, the, like, people's working memory on different patterns and see how much can the, can the neural network explain the data? And then we compare our cognitive model. It's more complex than the typical one, but it's still a cognitive model in and fully interpretable.

Integrating AI Insights into Psychology Research

00:08:07 - it's, didn't do as well as neural network, but it's get close. So that's like, so we think, okay, that's good enough because it's already close to a neural network. I have this, idea or belief now that, there is something wrong or at least incomplete with the way how, psychology has been studying things because mainly because it's fragmented and,

00:08:38 - after 20 years we discover a lot of things and then people forget about those older things and then keep digging new things. We never evolve into overall framework. So that's why I think, okay, maybe we can borrow this insight from the Artificial intelligence and try to build something that's more complex and, but more complete. So for that purpose, that's why we need, large scale experiment.

00:09:09 - So with this general idea, I started with the topic I'm more familiar with, but now with that in hand now, with a bunch of collaborators, we are moving towards something that's more real world relevant, like also like, using large scale experiment. for example, in this one study, we, a, a few collaborators, collaborator, collaborators and I, we are trying to figure out what is,

Advancing Emotional Models in Psychology

00:09:44 - in our ma in our mind. you may have heard that, in psychology there is this model, on emotion. Some of them is divided into different categories like happiness or sadness, et cetera. Some of them like, map all those different emotions into in A 2D space. But, we tend to believe that neither of those are complete. And the reason why we don't have anything better,

00:10:17 - scale, high quality data for us to build a model on it. And, that's what we are doing now and the few other things on visual beauty and also on visual expression. And that's what, we like, I I just mentioned. And also on like, language understanding or the different things. So hopefully those will make, like, larger impact

00:10:50 - world applications. but we are working on that.

Faculti AI can make mistakes

Research

Policy

Analysis