1 |
Yo emacs, this document is -*-text-*-. |
2 |
|
3 |
ORDERING THE CHAOS IN PARTICIPANT SUBMITTED APPRAISALS |
4 |
|
5 |
The purpose of this design document is to give some structure to |
6 |
free-form appraisals. We want to tease out the implicit structure and |
7 |
parameterize it without accidentally creating obstacles to creative |
8 |
and varied appraisals. |
9 |
|
10 |
WLJ: I've often found that what seems obvious and true to me is not so |
11 |
obvious and true to another. By the time I've slogged through the |
12 |
whole thing with them, the original idea is often refined quite a bit. |
13 |
|
14 |
|
15 |
*** CONTEXT *** |
16 |
|
17 |
The representation of context is guided by Baron-Cohen's Theory of |
18 |
Mind. |
19 |
|
20 |
A good theory of mind would be able to generate believable affective |
21 |
evaluations from any perspective. Perspective (e.g. whether it is |
22 |
from Mommy's eyes, or Tracy's eyes, or the readers eyes) should be a |
23 |
parameter that delineates the range of sentences that comprise the |
24 |
items we use in our experiment. Perspective taking ability is a key |
25 |
function of the TOMM and is also what is believed to be impaired in |
26 |
Autistic Spectrum Disorders. |
27 |
|
28 |
** Associated with each cue is a list of participating intentional |
29 |
agents. Anthropomorphic agents are OK for inclusion. |
30 |
|
31 |
** appraising-agent: Who is appraising the situation? The resulting |
32 |
appraisal is the appraising-agent's opinion about or construal of the |
33 |
situation cue. |
34 |
|
35 |
Example: |
36 |
Goal: Eric wants to go in the car. |
37 |
Situation: Eric and his daddy are going on the train. |
38 |
Question: How will his daddy feel when he goes on the train? |
39 |
|
40 |
appraising-agent = his daddy |
41 |
|
42 |
** participating-agent: The other person (besides the |
43 |
appraising-agent) who is involved in the situation. Initially we are |
44 |
limiting appraisals to two people: the appraising-agent and the |
45 |
participating-agent. |
46 |
|
47 |
WLJ TODO, TERMINOLOGY CLARIFICATION: If we include the appraiser in |
48 |
the participating-agents relation then how do we quickly refer to the |
49 |
other non-appraiser participating-agent? |
50 |
|
51 |
Example: |
52 |
Goal: Eric wants to go in the car. |
53 |
Situation: Eric and his daddy are going on the train. |
54 |
Question: How will Eric feel when he goes on the train? |
55 |
|
56 |
appraising-agent = Eric |
57 |
participating-agent = Daddy |
58 |
|
59 |
Two people is the minimum number of people needed for TOM |
60 |
propositions. For our simple appraisal model, there are exactly two |
61 |
participants for any given appraisal. (Yet, there can be more than |
62 |
two participants per cue.) For any given appraisal, there is one |
63 |
appraising-agent. |
64 |
|
65 |
WLJ TODO: I want to parameterize one more facet of appraisal. Given |
66 |
this cue / appraisal: |
67 |
|
68 |
At bedtime daddy makes Toby some hot chocolate. |
69 |
participatingAgents = Toby, daddy |
70 |
appraiser = Toby |
71 |
|
72 |
Which parameter distinguishes between: |
73 |
|
74 |
How does Toby imagine Toby feels about receiving hot chocolate? |
75 |
(or simply "How does Toby feel about receiving hot chocolate?") |
76 |
|
77 |
How does Toby imagine his daddy feels about giving hot chocolate? |
78 |
|
79 |
In other words, who is the appraiser trying to mindread? Can we call |
80 |
this parameter "point of view" or "mindreadee" or do you have a better |
81 |
idea? |
82 |
|
83 |
** Context Summary |
84 |
|
85 |
+ Each cue is associated with a list of participanting intentional |
86 |
agents. |
87 |
|
88 |
Each appraisal has: |
89 |
|
90 |
+ Two participants. |
91 |
|
92 |
+ One of the participants is the appraising-agent. |
93 |
|
94 |
+ The appraiser is trying to mindread either him/herself or the other |
95 |
participant. |
96 |
|
97 |
|
98 |
*** GOAL EVALUATION *** |
99 |
|
100 |
Goal, AntiGoal, and NoGoal are *relationships* that agents have with |
101 |
respect to certain states of affairs. |
102 |
|
103 |
unfulfilled fulfilled |
104 |
----------- -------------- |
105 |
Goal happy |
106 |
AntiGoal sad |
107 |
NoGoal indifferent |
108 |
|
109 |
G-type English sentence template |
110 |
----------- ---------------------------- |
111 |
Goal I want ... |
112 |
AntiGoal I don't want ... |
113 |
NoGoal I am indifferent about ... |
114 |
|
115 |
** Basic examples: |
116 |
|
117 |
Goal exmaple: Eric wants to go on the train. Eric and his daddy are |
118 |
going on the train. |
119 |
|
120 |
AntiGoal Example: Billy doesn't want a diagonally cut |
121 |
sandwich. (statement of anti-goal). Mommy gives Billy such a |
122 |
sandwich. (emotion eliciting situation) Billy cries. (If an |
123 |
AntiGoal is met and Billy is sufficiently bratty and inflexible then |
124 |
he cries.) |
125 |
|
126 |
The relation between an infant and "Wesley Clark wins Democratic |
127 |
Nomination" is the NoGoal relation. (This is true bc infants |
128 |
have no cognitive capacity for an understanding of politics.) |
129 |
|
130 |
** Long-term examples: |
131 |
|
132 |
Most people have a long-term goal like: "I want to be successful in my |
133 |
life." For a person P, the relationship between P and "P is |
134 |
successful" is the Goal relation. |
135 |
|
136 |
A no-goal example: "I have no grand plan for what I want to do with my |
137 |
life." |
138 |
|
139 |
There exist long term anti-goals that can be associated with a |
140 |
suicidal outlook. One example of such a long term anti-goal is "Fred |
141 |
has ``Fred lives a long life'' as an anti-goal." |
142 |
|
143 |
There are also long term anti-goals that are the opposite of suicidal. |
144 |
For example "Bob has ``Bob dies early''" as an anti-goal. |
145 |
|
146 |
Another example of a long-term anti-goal is: "I don't want to pursue |
147 |
the same career as my father." |
148 |
|
149 |
Usually when I talk about goals, I am implicitly talking about |
150 |
short-term goals. I usually assume "normal" or common-sense |
151 |
long-term goals. |
152 |
|
153 |
** Attractiveness |
154 |
|
155 |
> but for the |
156 |
> sake of discussion we can assume an equivalence: it desirable to be |
157 |
> appealing and undesirable to be unappealing. |
158 |
|
159 |
Yes. Kind of puzzling that part of Ortony. But I suppose it what makes |
160 |
it a powerful theory? It sorta kinda feels like Freudian theory in that |
161 |
you can view somethign as just a cigar or as something that is more than a |
162 |
cigar. I'm speaking loosely and metaphorically. In spite of the |
163 |
reference to Freud, I do think there is something scientifically testable |
164 |
in OCC. |
165 |
|
166 |
TODO for Joshua: propose definition of attractiveness in terms of |
167 |
goals. c/c with Ortony |
168 |
|
169 |
|
170 |
** KR Implementation |
171 |
|
172 |
WLJ: I am interested in the slots that instances of Goal-Jarrold might |
173 |
have. My KM model, offers some glimmers in this regard. Much much |
174 |
more work needs to be done in this area. Tested working rules which |
175 |
determine the degree to which a Goal-Jarrold or a Individual-Intention |
176 |
has been met/violated/etc are essential drivers of this part of this |
177 |
region of the ontology. |
178 |
|
179 |
|
180 |
** Personality |
181 |
|
182 |
Adding "personality," that would be a fun and interesting direction to |
183 |
take the model. Then some situations would be in the form of ... |
184 |
|
185 |
{Tracy/Eric/Toby} wants a(n) {apple/banana/hot chocolate/ride in a car/ |
186 |
ride in a train}. (He/she) is a {good kid/shy kid/bratty |
187 |
kid/firestarter}. |
188 |
|
189 |
...with a little creativity we can add some interesting |
190 |
parameters to the overall model that do not overly tax |
191 |
the KR abilities of our system yet do cause our scenarios |
192 |
for people to rate to be more interesting. |
193 |
|
194 |
|
195 |
*** AFFECTIVE STATES *** |
196 |
|
197 |
There are particular affective states which are triggered when a |
198 |
fulfilled Goal changes in type (from Goal to AntiGoal, etc). |
199 |
Investigate this later. |
200 |
|