CBB 2012: Metamemory and evolutionary dynamics in cognitive processes

Transcript

1. METAMEMORY & evolutionary dynamics in cognitive processes by Jordan Suchow
2. <refrain>
4. metamemory
5. visual working metamemory
6. it reflects both the existence and the precision of the underlying memories
7. and people can use it to guide their selection of an item
9. metamemory poses a problem for standard models of working memory
10. random process of degradation
11. {re}allocation
12. </refrain>
13. <1>
14. metamemory
15. what’s that? 1
16. “I don’t remember.” 1
17. “I’m not sure.” 1
18. “Can you remind me of something?” 1
19. “I can’t remember jack.” 1
21. $ 1
23. 1 standard visual working memory task:
24. 600 ms Encoding 900 ms Retention Response screen 1
25. +
26. +
27. +
28. +
29. + error (deg)
30. 1 metamemory task #1:
31. +
32. +
33. +
34. +
35. +
36. +
37. 1 error (deg)
38. 1 (interleaved in random order.)
39. +
40. +
41. +
42. 1 either:
43. +
44. 1 or:
45. +
46. 1 the data:
47. −3 −2 −1 0 1 2 3 0 100 200 300 400 500 1 1 error (deg) 0 +180 -180 precision guess rate
48. 1 1 precision guess rate 1 2 0 0.05 0.1 0.15 0.2 0.25 ` 0% 10% 20% random best random best 2 items 1 item 3 items 1 2 0 2 4 6 8 10 12 14 16 18 20 0° 10° 15° 5°
49. visual working metamemory exists
50. it reflects both the existence and the precision of the underlying memories
51. and people can use it to guide their selection of an item
52. a second metamemory task: 1
53. 600 ms Encoding 900 ms Retention Response screen 1, confidence ratings (arcs)
54. +
55. +
56. +
57. +
58. +
59. +
60. the data: 1, confidence ratings (arcs)
61. confidence rating (size of drawn arc, in degrees) 1, confidence ratings (arcs)
62. performance on the task (error, in degrees) 1, confidence ratings (arcs)
63. −3 −2 −1 0 1 2 3 0 100 200 300 400 500 1 error (deg) 0 +180 -180 precision guess rate 1, confidence ratings (arcs)
64. 1 precision guess rate 1 2 3 0 0.1 0.2 0.3 0.4 1, confidence ratings (arcs) 1 2 3 0 5 10 15 20 25 small arc medium arc LARGE ARC small arc medium arc LARGE ARC
65. visual working metamemory exists
66. it reflects both the existence and the precision of the underlying memories
67. where does it come from? 1
68. let’s rule out some of the more possibilities: 1
69. were some colors just easier to remember than others? 1
70. (no.) 1 14.5° 14.6° 13.5° 14.2° 13.1° 14.3°
71. were some locations just easier to remember than others? 1
72. 1 (no.) 19.6° 21.5° 21.1°
73. were some items allotted more of the resource than others? 1/6 1/3 1/2 1
74. 1/6 1/3 1/2 1
75. approach: look for tradeoffs 1
76. +
77. +
78. +
79. +
80. +
81. +
82. approach: look for tradeoffs 1
83. 1 when 2nd response is good: when 2nd response is bad: precision of first response...
84. 1 when 2nd response is good: 26.0° when 2nd response is bad: precision of first response...
85. 1 when 2nd response is good: 26.0° when 2nd response is bad: 25.6° precision of first response...
86. (same deal for all pairwise comparisons.) 1
87. what does this all say about visual working memory?
88. visual working metamemory exists
89. it reflects both the existence and the precision of the underlying memories
90. people can use it to guide their selection of an item
91. and, importantly, they aren’t relying on differences in allocation, nor on the memorability of certain colors or locations
92. </1>
93. <refrain>
95. metamemory
96. visual working metamemory
97. it reflects both the existence and the precision of the underlying memories
98. and people can use it to guide their selection of an item
100. metamemory poses a problem for standard models of working memory
101. random process of degradation
102. {re}allocation
103. </refrain>
104. <2>
105. metamemory poses a problem for standard models of working memory
106. model 1: a continuously-divisible resource, divided evenly among items 2
107. 1 2
108. 1/2 1/2 2
109. 1/3 1/3 1/3 2
110. 1/3 1/3 1/3 uh oh. 2
111. fail. (can’t account for metamemory data.)
112. model 2: a discrete “slot”-based resource, divided evenly among items 2
113. each participant has a fixed number of slots 2
114. 4 2
115. 1 <slot> <slot> <slot> <slot> 2
116. 1 <slot> <slot> <slot> <slot> 2
117. 1 <slot> <slot> <slot> <slot> 2
118. 1/4 1 <slot> <slot> <slot> <slot> 2
119. 0 2 1/4 1 <slot> <slot> <slot> <slot>
120. 0 0 2 1/4 1 <slot> <slot> <slot> <slot>
121. metamemory under this model? 2
122. 3 cases: 2
123. # of slots < n <slot> <slot> 0
124. (# of slots ≥ n) && (evenly divisible) <slot> <slot> <slot>
125. (# of slots ≥ n) && (~evenly divisible) <slot> <slot> <slot> <slot>
126. uh oh. 2
127. improve only in precision XOR guess rate 2
128. fail. (can’t account for metamemory data.)
129. 2 we need a new framework.
130. the framework: a random process of degradation 2
131. 1/3 1/3 1/3 2
132. 1/4 1/3 1/3 2
133. 1/6 1/5 1/3 2
134. RIP 1/3 2 1/6
135. RIP 1/3 2 1/6 yay.
136. allows for the metamemory results.
137. two particular instantiations, borrowed from the field of population genetics: 2
138. 2, pure death process time
139. time RIP RIP RIP 2, pure death process
140. time RIP RIP RIP 2, pure death process
141. time RIP RIP RIP 2, pure death process
142. time RIP RIP RIP 2, pure death process
143. time RIP RIP RIP 2, pure death process
144. time RIP RIP RIP 2, pure death process
145. time RIP RIP RIP 2, pure death process
146. time RIP RIP RIP 2, pure death process
147. time RIP RIP RIP 2, pure death process
148. time RIP RIP RIP 2, pure death process
149. time RIP RIP RIP 2, pure death process
150. Time step 1 102 Time step 1 10 100 1 Number of stored items n=5 n=3 n=1 2, pure death process
151. unlike real memory data 2, pure death process
152. how can we keep the representations alive? 2
153. {re}allocation 2, k-allele Moran process
154. time A B C D E F G H I t = 0 t = 1 t = 2 t = 3 t = 4 t = 5 C A B D E F G H I A B D E F G H I G B D G H I E I A G D G H I E I G D B D G H I E I G D I 2, k-allele Moran process A B D E F G H I C
155. time A B C D E F G H I t = 0 t = 1 t = 2 t = 3 t = 4 t = 5 C A B D E F G H I A B D E F G H I G B D G H I E I A G D G H I E I G D B D G H I E I G D I 2, k-allele Moran process
156. time A B C D E F G H I t = 0 t = 1 t = 2 t = 3 t = 4 t = 5 C A B D E F G H I A B D E F G H I G B D G H I E I A G D G H I E I G D B D G H I E I G D I 2, k-allele Moran process
157. time A B C D E F G H I t = 0 t = 1 t = 2 t = 3 t = 4 t = 5 C A B D E F G H I A B D E F G H I G B D G H I E I A G D G H I E I G D B D G H I E I G D I 2, k-allele Moran process
158. time A B C D E F G H I t = 0 t = 1 t = 2 t = 3 t = 4 t = 5 C A B D E F G H I A B D E F G H I G B D G H I E I A G D G H I E I G D B D G H I E I G D I 2, k-allele Moran process
159. time A B C D E F G H I t = 0 t = 1 t = 2 t = 3 t = 4 t = 5 C A B D E F G H I A B D E F G H I G B D G H I E I A G D G H I E I G D B D G H I E I G D I 2, k-allele Moran process
160. time A B C D E F G H I t = 0 t = 1 t = 2 t = 3 t = 4 t = 5 C A B D E F G H I A B D E F G H I G B D G H I E I A G D G H I E I G D B D G H I E I G D I 2, k-allele Moran process
161. 100 102 0 2 4 6 1 10 100 1000 Time step Number of recalled items n=7 n=5 n=3 n=1 2, k-allele Moran process
162. a random process of degradation can account for metamemory
163. reallocation keeps memories alive for a long time
164. these dynamics might account for capacity limitations
165. </2>
166. <refrain>
168. metamemory
169. visual working metamemory
170. it reflects both the existence and the precision of the underlying memories
171. and people can use it to guide their selection of an item
173. metamemory poses a problem for standard models of working memory
174. random process of degradation
175. {re}allocation
176. </refrain>
177. by Jordan Suchow thanks.
178. by Jordan Suchow thoughts?