PAN Lab
Join the advisory group for PAN Lab teams
I am Stephen Lieberman, and I build the PAN Lab. PAN Lab teams, which I am building now, will let people build and test a model of an AI deployment together. A team can bring together the organization that runs a system, the frontline staff who use it, and the people that the system makes decisions about.
I am forming an advisory group to help judge whether the teams system hears every member, and whether each member's experience shapes the model that their team builds. This page says who I am looking for, what the group would do, and how to tell me you are interested.
I am building the teams system now, as part of PAN Lab v0.1, an early prototype.
Why this group
A model that a group builds can use only what its members put into it. A review of 65 studies of group decision making (Lu, Yuan, and McLeod, 2012) found that groups tend to discuss what every member already knows, and to pass over what only one member knows. In a team that models an AI deployment, the member with that knowledge may be someone whose case the system decided, or a worker who saw the system fail.
I have designed the teams system to record what each member contributes, and to show the team which of those contributions its model uses. That design is untested. The advisory group's job is to help me test it honestly.
What the evaluation can show, and what it cannot
The evaluation can show whether members use these features, whether their team's model uses what they contribute, and whether they report feeling heard. It cannot show that the features make better models or better governance, and I will not claim that it does.
Who I am looking for
The group needs people who know these systems from different sides:
- People with lived experience of harm from an AI system used in a public service, such as benefits, housing, child welfare, or the courts. The group will include at least one such person.
- Frontline practitioners who work with the people whose cases these systems decide, such as caseworkers, social workers, and benefits staff.
- People from the organizations that deploy these systems, who know how an organization decides to use one.
You do not need any technical background. If you study these systems, or come to this work another way, you are welcome to write too.
What members would do
- Read short, anonymized summaries of how teams used the system, built only from the records of members who agreed to research use.
- Meet by video to say where members seem to be heard, and where they do not.
- Say what to change and what to remove, including whether a feature should go, or whether the system should not be used for some purpose at all.
- Read how the evaluation reports its findings before anything is published, and say where it claims too much.
Time and pay
I expect the work to need about two hours a month while the prototype is evaluated: one video meeting of about an hour, and about an hour of reading before it. You can miss a meeting, or leave the group, at any time and without giving a reason.
The group is unpaid while the teams system is a prototype.
What happens to what you send
Only the site's administrators can see what you send, and I will use your email address only to write to you about the advisory group.
Nothing you share, in this form or in the group, is published without your consent.
You can withdraw at any time. The confirmation email contains a link that deletes your answers, or you can write to Stephen@Paramerge.com and I will delete them.
Tell me you are interested
The form takes a few minutes. You can leave any field marked optional blank.