luxzia/common_tasks_and_XAI.csv

## common_tasks_and_XAI.csv

          
            task
            users who many engage in this task
            example questions they may ask the AI

            
              to improve or debug the model
              Model Developers. Some applications would also allow other user groups to perform this task.
              1. Is the performance of the AI good enough? 2. How does the AI make predictions? How might it go wrong? 3. Why does the AI make such a mistake?

            
              to evaluate the capability of the AI and form appropriate trust
              All user groups can engage in this task at some point
              1. Is the performance of the AI good enough? What are the risks and limitations? 2. What kinds of output can the AI give? 3. How does the AI work? Is is reasonable?

            
              to make informed decisions or take better actions?
              Decision-Makers. Impacted Groups. More.
              1. Why is this instance predicted to be X? 2. Why is this instance not predicted to be Y? 3. How to change this instanceto be predicted Y? 4. How to make sure this instance remains to be X? What change is permitted?

            
              to adapt usage or control
              Decision-Makers. Business Owners. And more.
              1. How does the AI make predictions? What canI supply or change for it to work well? 2. What if I make this change?

            
              to learn about new knowledge about a domain
              Decision-Makers. Business Owners. Impacted Groups. And More.
              1. How does the prediction task work? What are the key features to consider? 2. What if this feature changes? How does it impact the outcome? 3. Why is this instance not predicted to be Y as I would expect?

            
              to ensure ethical or legal compliance
              all user groups can engage in this task at some point
              1. how does the AI make predictions? Are there any legal or ethical concerns such as discrimination or privacy or security concerns? 2. Why are the two instances or groups not treated the same by the AI?
task	users who many engage in this task	example questions they may ask the AI
to improve or debug the model	Model Developers. Some applications would also allow other user groups to perform this task.	1. Is the performance of the AI good enough? 2. How does the AI make predictions? How might it go wrong? 3. Why does the AI make such a mistake?
to evaluate the capability of the AI and form appropriate trust	All user groups can engage in this task at some point	1. Is the performance of the AI good enough? What are the risks and limitations? 2. What kinds of output can the AI give? 3. How does the AI work? Is is reasonable?
to make informed decisions or take better actions?	Decision-Makers. Impacted Groups. More.	1. Why is this instance predicted to be X? 2. Why is this instance not predicted to be Y? 3. How to change this instanceto be predicted Y? 4. How to make sure this instance remains to be X? What change is permitted?
to adapt usage or control	Decision-Makers. Business Owners. And more.	1. How does the AI make predictions? What canI supply or change for it to work well? 2. What if I make this change?
to learn about new knowledge about a domain	Decision-Makers. Business Owners. Impacted Groups. And More.	1. How does the prediction task work? What are the key features to consider? 2. What if this feature changes? How does it impact the outcome? 3. Why is this instance not predicted to be Y as I would expect?
to ensure ethical or legal compliance	all user groups can engage in this task at some point	1. how does the AI make predictions? Are there any legal or ethical concerns such as discrimination or privacy or security concerns? 2. Why are the two instances or groups not treated the same by the AI?