Hi! 👋 Welcome to The Big Y!
Trust in AI is key in applications where there are human-to-AI interactions that have both positive or negative implications. The perception of a user interacting with an AI system can have a significant impact on the effectiveness of the system, thus for those developing AI systems, it is extremely important to continue to study and better understand how people perceive these systems.
For example, a study looking at the impact of AI feedback on employee productivity found an interesting impact of two countervailing effects. AI generated feedback helped increase employee job performance, with employees attaining 12.9% higher job performance when receiving AI feedback over those receiving human manager feedback. But, at the same time, the employees that were told they were receiving AI generated feedback had a 5.4% lower job performance than those who were told they were receiving feedback from human managers. If this sounds complicated it may help to take a look at the graph below.
An application where you can see some of this in action is when AI is used to provide automated feedback on homework or test assignments. At Stanford, one of the online computer science courses uses AI to deliver detailed feedback and critique of student tests and homework. The feedback from the AI is not always perfect, but according to Stanford, students agreed with the AI feedback 97.9% of the time, while only agreeing with human teacher feedback 96.7% of the time.
In this situation, AI opens up the door for better feedback for more students, but it only works if the students trust the feedback that is being given to them. This also means that the level of quality needs to be very high for the AI being used, as anything subpar would undermine trust.
AI used in hiring situations continues to be a point of contention and discussion. AI hiring games and analysis can be discriminatory against job seekers with disabilities. The problem is that the government body responsible for fighting employment discrimination has no guidelines to address potential issues.
Thanks for reading! Share this with a friend if you think they'd like it too. Have a great week! 😁
🎙 The Big Y Podcast: Listen on Spotify, Apple Podcasts, Stitcher, Substack