GD POLITICS Podcast Summary
Episode: California Gerrymandering, AI Polling, And Ukraine
Host: Galen Druke
Guests: Mary Radcliffe, Nathaniel Rakich
Date: August 18, 2025
Podcast: GD POLITICS
Overview
This episode of GD POLITICS dives into three major topics shaping U.S. politics:
- The intensifying "gerrymandering wars," with California and Texas leading a new round of partisan redistricting,
- The emergence of AI-driven polling techniques, particularly Ipsos and Stanford’s new digital twin respondent project,
- The evolving U.S. public attitudes toward the ongoing war in Ukraine, especially in light of recent meetings between President Trump, President Zelensky, and President Putin.
Hosts Galen Druke, Mary Radcliffe, and Nathaniel Rakich analyze and debate these subjects with their characteristic mix of data, skepticism, and humor.
Key Discussion Points and Insights
1. U.S. Attitudes on Ukraine, Trump, and Foreign Policy
[02:28 – 14:14]
-
Public Distrust in Trump on Russia/Ukraine
- Mary cites a Fox survey showing 58% of voters think Putin has the upper hand in Ukraine, only 35% chose Trump ([03:08]):
- "I don't think the public has a lot of confidence in Trump's dealing with Putin in general." — Mary Radcliffe [03:08]
- Nathaniel highlights Pew data: 33% think Trump would favor Russia too much, only 6% would favor Ukraine, 28% thought he’d strike the right balance, and 32% were unsure ([04:12]).
- "You do have to bear in mind with public opinion on foreign policy that it can be pretty fluid and uncertain." — Nathaniel Rakich [04:49]
- Mary cites a Fox survey showing 58% of voters think Putin has the upper hand in Ukraine, only 35% chose Trump ([03:08]):
-
Media Attention and Public Priorities
- Galen notes the lack of media focus on Trump and Russia in the second term compared to the first, despite the friendship-like body language at recent meetings with Putin ([06:01]).
- Mary and Nathaniel agree that foreign conflicts, including Ukraine, are not top of mind for the American public, with polling showing more interest in the economy and leadership ([06:01], [10:09]).
-
Opinion Dynamics and Republican Attitudes
- Mary explains, using Chicago Council data, how Republican support for Ukraine aid plummeted following Trump's changing stance, then slightly rebounded, while Democrats’ support remained higher but also declined over time ([06:43]).
- "People are responding to their trusted sources of news and information...they're following the lead of their political leaders." — Mary Radcliffe [07:57]
- Debate over whether these shifts reflect deep convictions or short-term elite cues ([08:32]).
- Mary explains, using Chicago Council data, how Republican support for Ukraine aid plummeted following Trump's changing stance, then slightly rebounded, while Democrats’ support remained higher but also declined over time ([06:43]).
-
Gallup Poll on National Priorities
- Only 2% see foreign conflicts as the top national issue. The economy is cited by 27%, but 26% now say “poor leadership”—a historically high number ([10:09]).
- Breakdown on the vagueness of “poor leadership” and whether this references Trump, other leaders, or a general frustration with government ([11:24]).
2. Gerrymandering Arms Race: Texas v. California
[14:14 – 34:25]
-
Retaliatory Gerrymandering: California’s Plan
- California, led by Governor Gavin Newsom, proposes a temporary suspension of its independent redistricting commission to add 5 Democratic districts, but only if Texas or another GOP state redraws first ([14:14]).
- The process requires a two-thirds legislative vote and a public referendum this fall ([15:48]).
-
Likelihood and Legal Hurdles
- Nathaniel doubts the likelihood of success, pointing to compressed timelines, legal uncertainty, and unpopularity of partisan map-drawing ([15:48], [16:15]).
- "There's real doubt about whether this will pass...and then there are also questions of legal challenges." — Nathaniel Rakich [16:15]
- Nathaniel doubts the likelihood of success, pointing to compressed timelines, legal uncertainty, and unpopularity of partisan map-drawing ([15:48], [16:15]).
-
Public Attitudes and Polling
- Americans overwhelmingly dislike gerrymandering: YouGov/Economist poll finds a 69–9 (60-point) margin against partisan map-drawing ([18:38]).
- "People really, really don't like gerrymandering." — Mary Radcliffe [18:38]
- Only 36% of Californians favor returning redistricting to lawmakers; by 2:1, they support the commission [20:45]. However, polling may shift after Newsom campaigns on this as a necessary response to GOP moves.
- Discussion of the complexities and biases in ballot measure polling ([22:56], [23:08]).
- Americans overwhelmingly dislike gerrymandering: YouGov/Economist poll finds a 69–9 (60-point) margin against partisan map-drawing ([18:38]).
-
National “Arms Race” Implications
- Other states (Ohio, Missouri, Indiana, possibly Florida) may jump into the fray, with GOP states having more leeway due to fewer legal restrictions ([29:11], [30:26]).
- Even Democratic-leaning states like New York are limited by their constitutions, creating a conservative advantage ([32:56]).
-
Potential for Federal Action
- Calls from some endangered/moderate Republicans and Democrats for a national gerrymandering ban ([31:36]).
- Skepticism that Congress will act while Trump and the current House leadership see partisan advantage in the status quo. Missed opportunity from the failure to pass a simpler ban under HR1 ([34:25], [35:47]).
3. AI and Synthetic Polling: Ipsos and Stanford's Project
[37:54 – 52:53]
-
What Is “Digital Twin” Polling?
- Mary explains: Ipsos intends to train AI models ("digital twins") on the behavior and responses of real panelists, then use the bots to answer survey questions in their stead ([37:54]).
- "They can use their Galen robot to find out what Galen thinks..." — Mary Radcliffe [38:32]
- Mary explains: Ipsos intends to train AI models ("digital twins") on the behavior and responses of real panelists, then use the bots to answer survey questions in their stead ([37:54]).
-
Is It Good Data, Bad Data—Or Not Data?
- Mary sees potential in modeling or low-stakes market research but draws a line at calling it true polling data for elections ([38:51]).
- Nathaniel is blunt: "This is not data. You're not talking to people." — Nathaniel Rakich [40:50]
-
Critiques and Limitations
- Accuracy is questionable—Stanford’s model only matched real responses at 85% accuracy; election margins often hinge on far smaller differences than that ([42:13]).
- Synthetic respondents tend to be more stereotypically partisan, losing the “unexpected” and nuance of real voters ([45:38], [47:38]).
- Both Mary and Nathaniel emphasize that you still need to re-poll real people to train and update the AI; thus, it won’t solve nonresponse bias ([49:58]).
- "Things change really fast...I don't know that these models are going to be able to do a good job of capturing those kinds of idiosyncrasies that we see in polling data all the time." — Mary Radcliffe [52:20]
-
Should Synthetic Data Be Included in Election Forecasts?
- Definitive answer: No—not for now ([51:17]).
Notable Quotes and Memorable Moments
- "People really, really don't like gerrymandering." — Mary Radcliffe [18:38]
- "This is not data. You're not talking to people." — Nathaniel Rakich [40:50]
- "If you force people to give a response, I bet you would get that 80, 20, Nathaniel. Like, people really, really hate this." — Mary Radcliffe [19:17]
- "People are responding to their trusted sources of news and information that may include the president pretty easily on this." — Mary Radcliffe [07:57]
- "If you send a bunch of journalists out to diners in Iowa...that's what we were doing for a lot of time and it was just very vibes based and not data." — Nathaniel Rakich [40:50]
- "Not right now. Not right now." — Mary Radcliffe on including synthetic polls in averages [51:17]
- "Things change really fast...people change their minds about stuff, and I don't know that these models are going to be able to do a good job capturing those idiosyncrasies." — Mary Radcliffe [52:20]
Timestamps for Important Segments
- Jokes about MSNBC, Clippy, and show open: [00:00–00:35]
- Trump, Putin & foreign policy polling: [02:28–14:14]
- Gerrymandering arms race: legal mechanisms & polling: [14:14–34:25]
- National implications & federal ban discussion: [31:36–35:46]
- AI polling debate (Ipsos/Stanford): [37:54–52:53]
- Show wrap-up: [52:53–53:00]
In Summary
This episode captures the evolving landscape of American electoral reform and political polling:
- The partisan battle over congressional maps has entered a new, openly retaliatory phase with both red and blue states seeking competitive advantage.
- Attempts to use AI as survey respondents may have potential in some areas but are roundly rejected as a reliable substitute for talking to real people—especially in contexts where small shifts in opinion matter.
- On Ukraine, public opinion is both skeptical of Trump and highly malleable, driven more by “vibes” and elite cues than deep convictions.
- Broad agreement exists among Americans that gerrymandering is undesirable, but partisan and institutional hurdles make reform difficult, and ironically, the spiral of redistricting may force the issue back to Congress.
The episode blends current events, polling data, methodological skepticism, and light banter, making it a rich guide to the latest in American political strategy and survey science.
