(1 year, 7 months ago)
Public Bill CommitteesQ
Anna Thomas: Yes, absolutely. In our model, we suggest that the impact assessment should incorporate not just the data protection elements, which we say remain essential, but equality of opportunity and disparity of outcome—for example, equal opportunity to promotion, or access to benefits. That should be incorporated in a model that forefronts and considers impacts on work.
Q
Anna Thomas: I would advise very clear additional rights, and a duty to notify in advance what, how and why AI is being used where it has these impacts, and where it meets the threshold that I was just asked about. I would also advise having more consultation throughout design, development and deployment, and ongoing monitoring, because AI changes, and there are impacts that we have not thought about or cannot ascertain in advance.
There should also be a separate obligation to conduct an algorithmic impact assessment. The Bill does nudge in that direction, but it says that there should be an assessment, rather than a data protection impact assessment. We suggest that the opportunity be grasped of clarifying that—at least in the workplace context, but arguably there are lessons more widely—the assessment ought to cover these fundamental aspects, and impacts at work.