Supply Side I: Politicians, Campaigns, and Elections

POLISCI 158 · AI & Politics · Session 8 · Monday, October 19, 2026

Simone Paci

Roadmap

  1. Part I: A conversation with Gail Pellerin
    • Explain how AI changes campaign power
  2. Part II: What does the evidence establish?
    • Distinguish persuasion from political behavior
  3. Part III: Evaluate a rule
    • Trace its mechanism, enforcement and trade-offs

What would you ask a legislator?

Take one minute. Write a question about an AI election rule.

  • What problem does it target?
  • How would it work?
  • What could it change?

Part I: A conversation with Gail Pellerin

Welcome, Assemblymember Gail Pellerin

California Assembly District 28

Chair, Assembly Elections Committee

Official California State Assembly portrait of Gail Pellerin

https://webapi.assembly.ca.gov/district-media/assets/members/assembly_member_28.jpg

Part II: What does the evidence establish?

Which outcome are we measuring?

Attitude

  • What someone reports believing or supporting

Political action

  • What someone does

Election result

  • What happens in an election

Teaching schematic

What did these field experiments find?

Typeset 17-word excerpt from Kalla and Broockman’s 2018 published abstract, attributed and labeled as an excerpt

https://doi.org/10.1017/S0003055417000363

What did the experiment compare?

Paid UK Prolific participants

Baseline attitudes → random assignment to petition and chat condition

Treatment

  • AI instructed to persuade the participant to sign the assigned petition

Control

  • Same chat interface; neutral-topic AI conversation

After the chat

  • Attitudes measured again
  • Opportunity to sign a real petition: name, email and completed submission

Teaching schematic · Hackenburg et al. (2026), working paper · https://arxiv.org/pdf/2604.09200

Can an AI conversation lead to action?

Original Figure 2 Study 1 panel, unchanged labels, units, estimates, confidence intervals and legend

Original Figure 2 Study 2 panel, unchanged labels, units, estimates, confidence intervals and axis

Figure 2; panels rearranged · https://arxiv.org/pdf/2604.09200

Part III: Evaluate a rule

How would this rule work?

  1. Alone · 1 min: choose one rule discussed today
  2. Pairs · 2 min: policy designer + implementation skeptic
  3. Share · 2 min: one mechanism, one limit

Target and mechanism

Enforcement

Trade-off

Solo or written answers welcome · Teaching schematic

What evidence would convince you a rule works?

Campaign power

Who gains capacity?

→

Measured outcome

What changes?

→

Policy mechanism

What could a rule change?

Teaching schematic

See you next class

Thank you, Assemblymember Pellerin

Wednesday, October 21

Post-Training and Reasoning Models · Yiqing Xu