Should We Be Nice to Bots?
This week, we talked with Project One Executive Director Chris Carroll who asked an intriguing question: “Should we be nice to bots?” With the recent news that UCSF will have its own enterprise version of OpenAI’s ChatGPT, we thought it would be a good time to spark discussion with our FAS colleagues.
Should we be nice to bots? I haven’t decided yet...
Chris Carroll
Executive Director Project One (Oracle ERP)
Enterprise Strategy & Transformation
Should We Be Nice to Bots?
During a recent AI-focused meeting, Project One Executive Director Chris Carroll asked an intriguing question: “Should we be nice to bots?” With the recent news that UCSF will have its own enterprise version of OpenAI’s ChatGPT, we thought it would be a good time to spark discussion with our FAS colleagues.
Tell us about your role.
I help oversee the implementation of Oracle Cloud ERP (Project One) from a business and user standpoint, meeting the university’s functional requirements. Project One aligns our Human Resources, Finance, Research Contracts and Grants, Supply Chain, and Student Information operations into one integrated enterprise resource planning (ERP) solution. (See the FAQs to learn more.)
Are you nice to bots?
I haven’t decided whether I’m “nice” to bots. Sometimes I’m intentionally polite, sometimes not. One reason I might be intentionally nice is that the model is learning, and I’d rather it learn pleasant discourse. I also know that being nice has a positive impact on one’s own mindset. One reason I don’t add extra politeness is the hidden energy cost behind every interaction.
Do you have a favorite AI prompt?
For fun, I like to ask it an Iron Chef-type question like “What can I make with...” and then I list ingredients that are in my kitchen. For work, I use AI primarily to synthesize complex documentation. Instead of reading everything or picking through using a keyword search, I ask: “Can Oracle do X?” or “How do these modules work together?” It gives me the pieces I need.
Oracle is rapidly adding native AI capabilities across all the Project One areas. When we go live, one AI feature available in the Project One Human Resource module is “check-ins” for performance management. After a meeting with an employee, a manager can enter a few notes. At the end of the year, AI aggregates them into a suggested narrative. It won’t score employees, but it helps avoid recency bias and helps balance the view of the full year. That’s the key benefit: more even, fair evaluations across the full 12 months.
FAVORITE COOKIE JAR TREAT
Classic chocolate chip cookie
Have a good story idea? Contact Kim LaPean and sign up to receive the latest One Good Thing.