Do those that said yes to (1) uncertainty; (2) lack of incentives; and (3) leadership has principles impact what kind of actions are they taking if any?
What types of challenges have you or your team faced in regards to using gen AI responsibly?
http://courses.atlas.illinois.edu/spring2016/STAT/STAT200/RProgramming/RegressionFactors.html
| action | sig | oddsratio | ci_low(2.5) | ci_high(97.5) |
|---|---|---|---|---|
| The text of the action | The significance of the test (*** = p < .001, NS = Not significant) |
Odds ratio. See note below about interpretation. | Lower limit of confidence interval for odds ratio | Upper limit of confidence interval for odds ratio |
Click here for interpretation help
Interpretation: A value in the oddsratio column indicates that participants were who said “Yes” to the text in the tab header were X times more likely to engage in the behavior in the action column.
Example:
Tab: (1) Said “Yes” to “There is a lack of clarity on what ‘responsibility’ looks like”
| action | sig | oddsratio | ci_low(2.5) | ci_high(97.5) |
|---|---|---|---|---|
| Want a grilled cheese | *** | 2.01 | 1.75 | 2.75 |
Participants who said that there is a a lack of clarity on what ‘responsibility’ looks like are 2 times more likely to want a grilled cheese than those who said there is not a lack of clarity on what ‘responsibility’ looks like.
(1) Said “Yes” to “There is a lack of clarity on what ‘responsibility’ looks like”
| text | odds | asterix | lower_ci | upper_ci |
|---|---|---|---|---|
| Work with colleagues who have job responsibilities are related to supporting AI fairness or responsibility | 1.2759 | NS | 0.7391 | 2.241 |
| Take ethical / responsible AI trainings | 0.9437 | NS | 0.5528 | 1.621 |
| Conduct audits of the gen AI tools | 1.3616 | NS | 0.7326 | 2.654 |
| Conduct fairness / bias testing | 2.8708 | *** | 1.4790 | 6.041 |
| Consider data privacy implications and take actions to protect data privacy | 1.1837 | NS | 0.6946 | 2.031 |
| Ask about the data or model to understand potential limitations or issues it has | 1.8000 | *** | 1.0215 | 3.270 |
| Conduct adversarial testing or red teaming (pretend to be a bad actor to test how a product we develop could be used for harm by bad actors | 1.3991 | NS | 0.6450 | 3.385 |
| Use explainability methods (e.g., to enable team members to better understand and probe the model) | 1.2565 | NS | 0.7030 | 2.319 |
| 12 Build transparency approaches (e.g., use documentation that can make dataset and model decisions transparent to others) | 1.7077 | NS | 0.9360 | 3.258 |
| No actions taken (to my knowledge) | 1.1335 | NS | 0.5869 | 2.315 |
(2) Said “Yes” to “There is a lack of incentives”
| text | odds | asterix | lower_ci | upper_ci |
|---|---|---|---|---|
| Work with colleagues who have job responsibilities are related to supporting AI fairness or responsibility | 1.9282 |
|
1.0869 | 3.437 |
| Take ethical / responsible AI trainings | 2.5414 |
|
1.4271 | 4.605 |
| Conduct audits of the gen AI tools | 3.9472 |
|
2.1667 | 7.223 |
| Conduct fairness / bias testing | 1.8115 |
|
0.9961 | 3.260 |
| Consider data privacy implications and take actions to protect data privacy | 1.5720 | NS | 0.8888 | 2.805 |
| Ask about the data or model to understand potential limitations or issues it has | 2.6868 |
|
1.5095 | 4.842 |
| Conduct adversarial testing or red teaming (pretend to be a bad actor to test how a product we develop could be used for harm by bad actors | 2.3065 |
|
1.1051 | 4.654 |
| Use explainability methods (e.g., to enable team members to better understand and probe the model) | 1.7748 | NS | 0.9764 | 3.192 |
| 12 Build transparency approaches (e.g., use documentation that can make dataset and model decisions transparent to others) | 1.5881 | NS | 0.8705 | 2.858 |
| No actions taken (to my knowledge) | 0.3182 |
|
0.1069 | 0.765 |
(3) Said “Leadership that has expressed commitment to responsible AI”
| text | odds | asterix | lower_ci | upper_ci |
|---|---|---|---|---|
| Work with colleagues who have job responsibilities are related to supporting AI fairness or responsibility | 2.8907 | *** | 1.7993 | 4.6841 |
| Take ethical / responsible AI trainings | 3.8870 | *** | 2.4085 | 6.3492 |
| Conduct audits of the gen AI tools | 3.7284 | *** | 2.1812 | 6.4921 |
| Conduct fairness / bias testing | 2.3777 | *** | 1.4445 | 3.9425 |
| Consider data privacy implications and take actions to protect data privacy | 1.8737 | *** | 1.1818 | 2.9865 |
| Ask about the data or model to understand potential limitations or issues it has | 2.5422 | *** | 1.5828 | 4.1128 |
| Conduct adversarial testing or red teaming (pretend to be a bad actor to test how a product we develop could be used for harm by bad actors | 2.2222 | *** | 1.1599 | 4.3345 |
| Use explainability methods (e.g., to enable team members to better understand and probe the model) | 2.3029 | *** | 1.4017 | 3.8092 |
| 12 Build transparency approaches (e.g., use documentation that can make dataset and model decisions transparent to others) | 3.3064 | *** | 2.0004 | 5.5314 |
| No actions taken (to my knowledge) | 0.2876 | *** | 0.1398 | 0.5517 |
Contingency tables
## $`Work with colleagues who have job responsibilities are related to supporting AI fairness or responsibility`
## # A tibble: 2 × 3
## text_challenges_clarity `Said 'No'` `Said 'Yes'`
## <fct> <int> <int>
## 1 Said 'No' to 'There is a lack of clarity on what 'responsibility' looks like' 46 25
## 2 Said 'Yes' to 'There is a lack of clarity on what 'responsibility' looks like' 137 95
##
## $`Take ethical / responsible AI trainings`
## # A tibble: 2 × 3
## text_challenges_clarity `Said 'No'` `Said 'Yes'`
## <fct> <int> <int>
## 1 Said 'No' to 'There is a lack of clarity on what 'responsibility' looks like' 40 31
## 2 Said 'Yes' to 'There is a lack of clarity on what 'responsibility' looks like' 134 98
##
## $`Conduct audits of the gen AI tools`
## # A tibble: 2 × 3
## text_challenges_clarity `Said 'No'` `Said 'Yes'`
## <fct> <int> <int>
## 1 Said 'No' to 'There is a lack of clarity on what 'responsibility' looks like' 56 15
## 2 Said 'Yes' to 'There is a lack of clarity on what 'responsibility' looks like' 170 62
##
## $`Conduct fairness / bias testing`
## # A tibble: 2 × 3
## text_challenges_clarity `Said 'No'` `Said 'Yes'`
## <fct> <int> <int>
## 1 Said 'No' to 'There is a lack of clarity on what 'responsibility' looks like' 60 11
## 2 Said 'Yes' to 'There is a lack of clarity on what 'responsibility' looks like' 152 80
##
## $`Consider data privacy implications and take actions to protect data privacy`
## # A tibble: 2 × 3
## text_challenges_clarity `Said 'No'` `Said 'Yes'`
## <fct> <int> <int>
## 1 Said 'No' to 'There is a lack of clarity on what 'responsibility' looks like' 40 31
## 2 Said 'Yes' to 'There is a lack of clarity on what 'responsibility' looks like' 121 111
##
## $`Ask about the data or model to understand potential limitations or issues it has`
## # A tibble: 2 × 3
## text_challenges_clarity `Said 'No'` `Said 'Yes'`
## <fct> <int> <int>
## 1 Said 'No' to 'There is a lack of clarity on what 'responsibility' looks like' 51 20
## 2 Said 'Yes' to 'There is a lack of clarity on what 'responsibility' looks like' 136 96
##
## $`Conduct adversarial testing or red teaming (pretend to be a bad actor to test how a product we develop could be used for harm by bad actors`
## # A tibble: 2 × 3
## text_challenges_clarity `Said 'No'` `Said 'Yes'`
## <fct> <int> <int>
## 1 Said 'No' to 'There is a lack of clarity on what 'responsibility' looks like' 63 8
## 2 Said 'Yes' to 'There is a lack of clarity on what 'responsibility' looks like' 197 35
##
## $`Use explainability methods (e.g., to enable team members to better understand and probe the model)`
## # A tibble: 2 × 3
## text_challenges_clarity `Said 'No'` `Said 'Yes'`
## <fct> <int> <int>
## 1 Said 'No' to 'There is a lack of clarity on what 'responsibility' looks like' 52 19
## 2 Said 'Yes' to 'There is a lack of clarity on what 'responsibility' looks like' 159 73
##
## $`Build transparency approaches (e.g., use documentation that can make dataset and model decisions transparent to others)`
## # A tibble: 2 × 3
## text_challenges_clarity `Said 'No'` `Said 'Yes'`
## <fct> <int> <int>
## 1 Said 'No' to 'There is a lack of clarity on what 'responsibility' looks like' 55 16
## 2 Said 'Yes' to 'There is a lack of clarity on what 'responsibility' looks like' 155 77
##
## $`No actions taken (to my knowledge)`
## # A tibble: 2 × 3
## text_challenges_clarity `Said 'No'` `Said 'Yes'`
## <fct> <int> <int>
## 1 Said 'No' to 'There is a lack of clarity on what 'responsibility' looks like' 58 13
## 2 Said 'Yes' to 'There is a lack of clarity on what 'responsibility' looks like' 185 47
##
## $`Work with colleagues who have job responsibilities are related to supporting AI fairness or responsibility`
## # A tibble: 2 × 3
## text_orghave_incentives `Said 'No'` `Said 'Yes'`
## <fct> <int> <int>
## 1 Said 'No' to 'There is a lack of incentives' 155 89
## 2 Said 'Yes' to 'There is a lack of incentives' 28 31
##
## $`Take ethical / responsible AI trainings`
## # A tibble: 2 × 3
## text_orghave_incentives `Said 'No'` `Said 'Yes'`
## <fct> <int> <int>
## 1 Said 'No' to 'There is a lack of incentives' 151 93
## 2 Said 'Yes' to 'There is a lack of incentives' 23 36
##
## $`Conduct audits of the gen AI tools`
## # A tibble: 2 × 3
## text_orghave_incentives `Said 'No'` `Said 'Yes'`
## <fct> <int> <int>
## 1 Said 'No' to 'There is a lack of incentives' 196 48
## 2 Said 'Yes' to 'There is a lack of incentives' 30 29
##
## $`Conduct fairness / bias testing`
## # A tibble: 2 × 3
## text_orghave_incentives `Said 'No'` `Said 'Yes'`
## <fct> <int> <int>
## 1 Said 'No' to 'There is a lack of incentives' 177 67
## 2 Said 'Yes' to 'There is a lack of incentives' 35 24
##
## $`Consider data privacy implications and take actions to protect data privacy`
## # A tibble: 2 × 3
## text_orghave_incentives `Said 'No'` `Said 'Yes'`
## <fct> <int> <int>
## 1 Said 'No' to 'There is a lack of incentives' 135 109
## 2 Said 'Yes' to 'There is a lack of incentives' 26 33
##
## $`Ask about the data or model to understand potential limitations or issues it has`
## # A tibble: 2 × 3
## text_orghave_incentives `Said 'No'` `Said 'Yes'`
## <fct> <int> <int>
## 1 Said 'No' to 'There is a lack of incentives' 162 82
## 2 Said 'Yes' to 'There is a lack of incentives' 25 34
##
## $`Conduct adversarial testing or red teaming (pretend to be a bad actor to test how a product we develop could be used for harm by bad actors`
## # A tibble: 2 × 3
## text_orghave_incentives `Said 'No'` `Said 'Yes'`
## <fct> <int> <int>
## 1 Said 'No' to 'There is a lack of incentives' 215 29
## 2 Said 'Yes' to 'There is a lack of incentives' 45 14
##
## $`Use explainability methods (e.g., to enable team members to better understand and probe the model)`
## # A tibble: 2 × 3
## text_orghave_incentives `Said 'No'` `Said 'Yes'`
## <fct> <int> <int>
## 1 Said 'No' to 'There is a lack of incentives' 176 68
## 2 Said 'Yes' to 'There is a lack of incentives' 35 24
##
## $`Build transparency approaches (e.g., use documentation that can make dataset and model decisions transparent to others)`
## # A tibble: 2 × 3
## text_orghave_incentives `Said 'No'` `Said 'Yes'`
## <fct> <int> <int>
## 1 Said 'No' to 'There is a lack of incentives' 174 70
## 2 Said 'Yes' to 'There is a lack of incentives' 36 23
##
## $`No actions taken (to my knowledge)`
## # A tibble: 2 × 3
## text_orghave_incentives `Said 'No'` `Said 'Yes'`
## <fct> <int> <int>
## 1 Said 'No' to 'There is a lack of incentives' 189 55
## 2 Said 'Yes' to 'There is a lack of incentives' 54 5
##
## $`Work with colleagues who have job responsibilities are related to supporting AI fairness or responsibility`
## # A tibble: 2 × 3
## text_orghave_leadership `Said 'No'` `Said 'Yes'`
## <fct> <int> <int>
## 1 Said 'No' to 'Leadership that has expressed commitment to responsible AI' 126 52
## 2 Said 'Yes' to 'Leadership that has expressed commitment to responsible AI' 57 68
##
## $`Take ethical / responsible AI trainings`
## # A tibble: 2 × 3
## text_orghave_leadership `Said 'No'` `Said 'Yes'`
## <fct> <int> <int>
## 1 Said 'No' to 'Leadership that has expressed commitment to responsible AI' 126 52
## 2 Said 'Yes' to 'Leadership that has expressed commitment to responsible AI' 48 77
##
## $`Conduct audits of the gen AI tools`
## # A tibble: 2 × 3
## text_orghave_leadership `Said 'No'` `Said 'Yes'`
## <fct> <int> <int>
## 1 Said 'No' to 'Leadership that has expressed commitment to responsible AI' 151 27
## 2 Said 'Yes' to 'Leadership that has expressed commitment to responsible AI' 75 50
##
## $`Conduct fairness / bias testing`
## # A tibble: 2 × 3
## text_orghave_leadership `Said 'No'` `Said 'Yes'`
## <fct> <int> <int>
## 1 Said 'No' to 'Leadership that has expressed commitment to responsible AI' 138 40
## 2 Said 'Yes' to 'Leadership that has expressed commitment to responsible AI' 74 51
##
## $`Consider data privacy implications and take actions to protect data privacy`
## # A tibble: 2 × 3
## text_orghave_leadership `Said 'No'` `Said 'Yes'`
## <fct> <int> <int>
## 1 Said 'No' to 'Leadership that has expressed commitment to responsible AI' 106 72
## 2 Said 'Yes' to 'Leadership that has expressed commitment to responsible AI' 55 70
##
## $`Ask about the data or model to understand potential limitations or issues it has`
## # A tibble: 2 × 3
## text_orghave_leadership `Said 'No'` `Said 'Yes'`
## <fct> <int> <int>
## 1 Said 'No' to 'Leadership that has expressed commitment to responsible AI' 126 52
## 2 Said 'Yes' to 'Leadership that has expressed commitment to responsible AI' 61 64
##
## $`Conduct adversarial testing or red teaming (pretend to be a bad actor to test how a product we develop could be used for harm by bad actors`
## # A tibble: 2 × 3
## text_orghave_leadership `Said 'No'` `Said 'Yes'`
## <fct> <int> <int>
## 1 Said 'No' to 'Leadership that has expressed commitment to responsible AI' 160 18
## 2 Said 'Yes' to 'Leadership that has expressed commitment to responsible AI' 100 25
##
## $`Use explainability methods (e.g., to enable team members to better understand and probe the model)`
## # A tibble: 2 × 3
## text_orghave_leadership `Said 'No'` `Said 'Yes'`
## <fct> <int> <int>
## 1 Said 'No' to 'Leadership that has expressed commitment to responsible AI' 137 41
## 2 Said 'Yes' to 'Leadership that has expressed commitment to responsible AI' 74 51
##
## $`Build transparency approaches (e.g., use documentation that can make dataset and model decisions transparent to others)`
## # A tibble: 2 × 3
## text_orghave_leadership `Said 'No'` `Said 'Yes'`
## <fct> <int> <int>
## 1 Said 'No' to 'Leadership that has expressed commitment to responsible AI' 142 36
## 2 Said 'Yes' to 'Leadership that has expressed commitment to responsible AI' 68 57
##
## $`No actions taken (to my knowledge)`
## # A tibble: 2 × 3
## text_orghave_leadership `Said 'No'` `Said 'Yes'`
## <fct> <int> <int>
## 1 Said 'No' to 'Leadership that has expressed commitment to responsible AI' 130 48
## 2 Said 'Yes' to 'Leadership that has expressed commitment to responsible AI' 113 12