Only 28% of AI end users report that transparency is good for them

Researchers did a survey of 400 end users, AI builders and policymakers about transparency in generative AI. Only 28% of end users report that present practice is good for them. For builders the rate is 53%, and for policymakers it is 49%. The authors report that the difference shows how control shapes what is made.

What they asked

The team did the survey and a group test with end users. All 3 groups agree on the meaning of transparency. They want precision and privacy.

The numbers

The difference between the importance of the transparency features and their availability is greater for end users than for builders. The features that persons rate first also change with the situation and with the structure of the task.

What is not known

The abstract does not tell how the 400 persons were selected or how many are in each group. It gives no data on the features that end users lack most. The authors want different transparency tools, but they test none.

Insight: New AI reporting rules serve institutions, not end users

A US group wants large AI developers to send a report each 3 months on AI research to a US office. A toolkit from Action State gives companies records to use if 2 sides disagree. Transparency is made for the parties with control, and end users get the minimum. It reads as a signal to monitor if any rule names end users.