-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathsudo-code.txt
49 lines (38 loc) · 2.1 KB
/
sudo-code.txt
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
Sample action sequence
step near, pick up kettle, pour water in cup, put away kettle, take out tea bag, put tea bag in cup, pick up spoon
mix tea, put away spoon , pick up cup, drink tea
--------------------------------------------------------
;available objects for the action
(chunk-type action name object1 object2 object3)
;chunks:
;actions, with slots that contain objects which can afford the action
;objects, state of the object---location, purpose, can they contain, or heat up things
;situation awareness, e.g., (water boiled y, water in mug y, tea in mug y)
;goal buffer:
;keep track of the current action
;need additional slot to keep track of the sequence?
;imaginal buffer:
;situation awareness---modify the state of the objects
;top down versus bottom up??
Declarative memory---what we expect the participant to already know; pre-knowledge
(mug, color yellow, location x, location y, purpose has sugar, state) ; modified in imaginal buffer, i.e., state empty
Fixed versus flexible sequences
- a state in the goal buffer that keeps track of the state of the task, and then uses the production to refer to the state
We can do this without using the visualization.
Example of an observer model:
- 'step near' is observed, model produces 'pick up kettle', 'take out tea bag'
Fixed sequences
- tea and water in mug before mixing tea
water boiled, water in mug, tea in mug ---> drink tea or mix tea and then drink tea
- productions are dependent on situational awareness and incoming information;
- productions that have a hierarchical structure: use high or low utility to achieve selection;
------------------------------------------
A simple way to construct the model:
- situation awareness (sa):
slot that keeps descriptions and states of the task
- for each step of the data, update sa
- then, given sa, we retrieve a chunk that represents the current phase and next phases
- and then we return (model output) the next steps in that chunk
- have additional bottom-up productions that do not change sa, but intervenes
(e.g., take spoon--mix)
- have bottom-up production that fires when a dangerous action is predicted