-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathcolumn_descriptions
48 lines (48 loc) · 2.77 KB
/
column_descriptions
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
init_choiceA: initial participant confidence in choice A
init_choiceB: initial participant confidence in choice B
init_choiceC: initial participant confidence in choice C
init_choiceD: initial participant confidence in choice D
init_time: time in seconds between loading question and submitting initial answer
adjusted_choiceA: adjusted participant confidence in choice A (after advice)
adjusted_choiceB: adjusted participant confidence in choice B (after advice)
adjusted_choiceC: adjusted participant confidence in choice C (after advice)
adjusted_choiceD: adjusted participant confidence in choice D (after advice)
adjusted_time: time in seconds between loading advice and submitting adjusted answer
question_num: question number, starting at 1
question: question text
optionA: choice A text
optionB: choice B text
optionC: choice C text
optionD: choice D text
correct_answer: correct answer letter (A, B, C, D)
justification: justification text
advice_answer: advised answer letter (A, B, C, D, E or None)
topic: MMLU topic category
participant_id: unique participant identifier
source: participant pool
advisor: identity of the advisor (AI chatbot, expert)
give_justification: whether a text justification was provided
score: participant score according to survey scoring mechanism
major: participant major(s), comma separated
time: time in seconds to submit the full survey
usage_description: free text response description of how participants used ChatGPT
heard_of: 1 if participants have heard of ChatGPT
used: 1 if participants have used ChatGPT before
used_in_class: 1 if participants have used ChatGPT in the classroom
answered_mc: 1 if participants have used ChatGPT on multiple choice questions
weight_on_advice: calculated weight on advice (see paper)
fam_topic: familiarity topic category
topic_familiarity: level of familiarity in text ("comfortable", "neutral", "uncomfortable")
net_familiarity: level of familiarity as an integer (-1, 0, 1, respectively)
uncomfortable: 1 if participants are uncomfortable in the topic area
neutral: 1 if participants feel neutral about the topic area
comfortable: 1 if participants are comfortable in the topic area
usage_level: sum of `heard_of`, `used`, `used_in_class`, `answered_mc`
question_group: `question_num` binned in groups of 5
question_id: unique question identifier
advice_is_correct: whether the `advice_answer` matches `correct_answer`
last_advice_is_correct: `advice_is_correct` for the same participant's last answer
correct_advice_count: count of correct advice for the same participant's previous questions
incorrect_advice_count: count of incorrect advice for the same participant's previous questions
init_advice_confidence: initial participant confidence in choice `advice_answer`
advice_confidence: adjusted participant confidence in choice `advice_answer`