All the Examples of Bad Thinking From Veep Season 1; I Found 29, GPT-4 Found 38
Learning from political satires what not to do
Issue No. 14
Veep (TV-MA) is a political satire about Selina Meyers—a senator-turned-vice-president who seems mired in misfortune.1 I spent my free time last week watching season 1 and noting down examples of bad thinking, something the show has no shortage of.
After doing a pass through the season, I had GPT-42 also do a pass through it. Between us, we found 60 unique examples of logical fallacies, cognitive biases, and loaded language. I found 29, GPT-4 found 38. There were only seven instances where we picked up on the same or similar thing, which means neither of us alone would have detected as many examples as we did working together.
I prompted it with the following:
I’ll give you the transcript for an episode of a TV show and I’d like you to give me a list of all examples of logical fallacies, cognitive biases, and loaded language, where the examples demonstrate how those things were used for political or personal gain. Give me all the instances, please. And no need to define each one, just the example, please. Ready?
I’d then give it the transcript3 for an episode. When responses weren’t clear, I’d ask one to two follow-up questions. Granted, the transcripts aren’t error-free—some lines have spelling mistakes, and most lines don’t have the character’s name in front of it, which can be confusing. I filtered out examples that didn’t make a whole lot of sense. They constituted roughly 15% of its responses.
I got progressively tired and occasionally distracted. It wasn’t the show that was tiring, but the act of simultaneously watching a show and doing mental lookups in realtime. Here’s how I did.
Detections over 8 episodes: [8, 4, 1, 4, 2, 3, 2, 5] = 29
Median: 3.5 detections
Mean: 3.6 detections
Standard deviation: 2.2 detections
GPT-4 did better.
Detections over 8 episodes: [4, 2, 4, 4, 7, 4, 6, 7] = 38
Median: 4 detections
Mean: 4.8 detections
Standard deviation: 1.8 detections
A constraint I became aware of was that I kept looking for examples that were top of mind. Put differently, I couldn’t get over my own availability bias (see issue 3 for more). If, say, the halo effect was on my mind, I was hypersensitive to that throughout an episode.
Some observations about GPT-4. It had a tendency to rely on implications for judgments. As in, a character might not have explicitly said something that would be indicative of a straw man fallacy, but their general demeanor might have implied they were prone to creating straw men arguments. Sometimes, when I probed, GPT-4 would give me dialogue to back up its judgment. Other times, it would admit there wasn’t a specific line it could share.
GPT-4 sometimes picked up on examples that were technically true, but not practically relevant, despite the prompt specifying that our goal was to find examples where something had been used for political or personal gain. For instance, it detected “There aren’t enough people to fill a canoe in here” as hyperbole, which it is, but in that particular scene, the line’s indicative of frustration more so than a calculated angle.
GPT-4 couldn’t always detect satire. In one scene, a staff member role-plays, demonstrating the sort of diatribe a critic might launch into. GPT-4 detected that as a genuine position the person was taking. Other times, it did an impressive job of detecting when a single word like “buddy” was being used disparagingly. That word shows up 17 times in four episodes in varying contexts.
GPT-4 is nondeterministic—the association between what you give it and what you get back isn’t guaranteed. Its overall voice changed on me midway. For the first few episodes, it broke each of its responses into three categories—logical fallacies, cognitive biases, loaded language—and would go on for a bit in describing each. Then the next morning, I gave it the exact same prompt for the rest of the episodes, and its responses were a single list and much more succinct.
Bad arguments and loaded language
Here’s a spelling out of 11 memorable examples from the show’s first season.
Appeal to emotion
A governor has just been traduced by the vice president. A hot mic caught her saying the governor wasn’t born in America and therefore couldn’t run for president. Midway through a televised interview, the governor drops a cynical line, pandering to an audience he knows well.
Governor Chung: … Proud enough to risk my life for this country. I’ve got a Purple Heart on my chest, but the one that beats inside of me is red, white, and blue.
Vice President (at the TV): No, he didn’t. He didn’t say that.
Governor Chung: And that’s why this is the greatest country on Earth.
Vice President: I am gonna throw up a leg. And then l’m gonna take that leg, and l’m gonna beat Scott … with it.
It’s complicated
Implying a situation is so complicated that it’s impossible to take a position on it can be an attempt to conceal an inconvenient truth (see issue 7). The vice president reaches for that to dismiss her daughter Catherine’s criticisms.
Catherine: You’re changing the name of a hurricane?
Vice President: Yeah. There was gonna be a Hurricane Selina. And that would have been a disaster for us.
Catherine: You are not Thor, Mom.
Vice President: No, Catherine, listen. It’s a much more complicated and nuanced situation.
Catherine: No, it’s really not.
Plausible deniability
Saying something without actually saying it gives a person an out if they’re ever cornered (see issue 4), as in this exchange.
Vice President: I mean, how are we gonna shut this negativity down? Amy, what are we gonna do?
Chief of Staff: Yeah, well, it’s inaccurate. Technically, the Secret Service guy was reassigned.
Staff: Yeah, and technically we didn’t put two oil guys on Clean Jobs Task Force.
And in this exchange, where the vice president is trying to get out of a possible scandal involving her being pregnant.
Chief of Staff: [W]e could say the pregnancy test was for me.
Vice President: Oh, you mean that would be why you reassigned the Secret Service guy? ‘Cause you were feeling too much pressure. Baby on board. That kind of thing? Is that what you mean?
Chief of Staff: Well, yeah. You can mention that.
Vice President: I mean, l can’t tell you what to do. It’s kind of a personal thing. But it’s really good to hear you’re thinking.
The vice president can’t tell her chief of staff what to do. Wink, wink.
Halo effect
The halo effect is when we form either a positive or negative overall impression of someone based on a single character trait or action, without evidence. For instance, someone looks good, so we assume they’re also kind. Or someone has a lot of money, so we assume they’re also intelligent.
The vice president is asked by her staff to wear glasses in order to look smart, presumably to be seen as authoritative. In a roomful of donors, the vice president abandons a small donor, but listens to a big one, presumably because his ideas must be good.
$50 donor: I made you a cake with your face on it.
Vice President (smiling): That looks like me after a long day in the Senate.
Chief of Staff (whispers in v-p’s ear): I think she’s only paid 50 bucks.
Vice President (abruptly): I have to go.
Platinum donor: Madam Vice President.
Vice President: Yes?
Chief of Staff: I’d like to introduce you to one of our actual platinum donors.
Platinum donor: I have some great ideas about tax reform I want to talk to you about.
Vice President: I want to hear them.
Appeal to consequences
An appeal to consequences is when we make a judgment about the truth of something that’s in front of us based on whether that thing’s consequences are favorable. In the show, it’s telling that a senator refuses to join the vice president’s clean jobs commission for fear of that upsetting members of the oil and plastics industry. The senator’s convictions aren’t driven by the truth, but by consequences.
Vice President: I wanted to talk to you about coming on board the clean jobs commission.
Senator: No, sorry, Selina … You [annoy] plastics, you [annoy] oil. And you do not want to [annoy] those guys.
Vice President: Barbara, are you really willing to sell your soul to the guys who make toothbrushes and... and... and... and... and the holders for toothbrushes?
Hindsight bias
The hindsight bias is when we look back at something, and genuinely believe that we could have perfectly predicted it. In this exchange between the vice president and a member of her staff, he doesn’t hesitate for a second to proclaim that a hat that made the vice president’s head look wonky was why she lost her run for president.
Vice President: What would you say were the two biggest campaign mistakes that we made?
Staff: You looked tired a lot and the hat.
Vice President: I liked the hat. What are you talking about?
Staff: The hat hurt us. Your head looked weird in the hat, that’s all I’m gonna say.
Vice President: Okay.
False analogy
A false analogy is assuming that because two things have one quality in common, then they must be generally alike. In one scene, we find the following exchange.
Chief of Staff: You’re pouring oil on clean jobs?
Vice President: Oh, please, please, Amy, grow up. This is how they solved the Cuban missile crisis. This is my Cuba.
Confirmation bias
Starting from a place of certainty and rejecting any evidence that disproves that conclusion is confirmation bias (see issue 1). As in this exchange between the vice president and her chief of staff, the moment the vice president learns of her approval rating.
Vice President: 66% disapprove? That’s everyone in America who’s awake right now.
Chief of Staff: Okay, let’s not concentrate on that today. Let’s concentrate on Ohio.
Vice President: How could that many Americans not like me? I think they got it wrong. They counted those numbers wrong or something. Or all those Americans are completely wrong.
Red herring
A red herring is a misdirection, a distraction. The vice president asks her team to release all their correspondence as a distraction tactic, diverting attention from specific issues by flooding the media with irrelevant information.
Chief of Staff: Okay, there's a White House request to publish all the Secret Service office personnel records.
Vice President: They want our records now? What are we gonna do? What?
Chief of Staff: Well, we have to release those. And that has got me thinking, why not release all of our … records? Full disclosure … publish everything.
Vice President: All of our emails, and all of our phone records.
Staff: Right, right, ‘cause they won’t have time to read everything. I mean, you can’t read everything. I don’t read half the stuff I’m supposed to.
Chief of Staff: So by showing that we have nothing to hide, then we can actually hide some stuff.
Appeal to authority
An appeal to authority is tying the truth of something to an important person’s sway over it. As in the following exchange between the vice president’s staff and the White House liaison.
Jonah: White House says we need to majorly redact this, okay?
Mike (Vice President’s staff): I’ll redact your … face.
Jonah: It doesn’t come from me, okay, so just do it.
Ad hominem
An ad hominem attack is when one goes after a person instead of the issue being discussed. In one scene, rather than the vice president reacting to the content of what a senator shared about why she didn’t want to be on the clean jobs commission, the vice president instead resorts to personal attacks.
Staff: How was [Senator] Hallowes?
Vice President: Oh, you know, the usual. (Mocking accent) “Selina, hi. I’m melting. I’m melting.” She is such a [expletive].
And here are all 60 examples broken up by episode.
Episode 1 · Fundraiser, 12 examples
I spotted 8 examples, GPT-4 spotted 4 examples (*), and we had no overlaps.
Ad hominem*: Amy’s comment about Dan, calling him a “massive and total [expletive]”, attacks his character rather than addressing a specific behavior or action.
Appeal to authority*: Jonah’s request to redact parts of the vice president’s speech doesn’t give a reason other than the White House asking for that—“It doesn’t come from me, okay, so just do it.”
Appeal to consequences: Senator Hallowes doesn’t want to join the vice president’s task force, because it would lead to upsetting members of the oil and plastics industry. Her convictions are informed not by the truth, but by consequences.
Appeal to flattery: Dan, hoping for a position on the vice president’s staff, gives her advice and then ends it with, “Hey, you’re the old pro here.”
False analogy: The vice president says of her clean jobs task force, “This is my Cuba.”
False dilemma*: When discussing the clean jobs commission, it’s implied that you either support the plastics industry or you are against it, ignoring other potential positions or solutions (see issue 6 for more on false dilemmas).
False precision: Dan, hoping for a position on the vice president’s staff, confidently states two things she did wrong that cost her the presidency.
Halo effect: The vice president is asked to wear a scarf to look pretty, and later on, glasses to look smart.
Hindsight bias: A member of the vice president’s staff confidently tells her why she lost her presidential bid—she looked tired a lot and her hat made her head look weird.
Humor effect: The vice president has a humorous catchphrase for her clean jobs initiative—“You know what I say, it’s a dirty job and I love to do it.” (The bias being that it’s easier to recall funny information.)
In-group bias*: The vice president and her chief of staff discuss Senator Hallowes and other senators in a way that suggests a lack of respect or consideration for those outside their immediate group. Their dialogue often includes derogatory comments and dismissals.
Red herring: Dan proposes blaming Senator Hallowes for an offensive joke that the vice president said publicly, on account of her hearing it from the senator privately.
Episode 2 · Frozen Yoghurt, 6 examples
I spotted 4 examples, GPT-4 spotted 2 examples (*), and we had no overlaps.
Ad hominem: A member of staff makes a suggestion to visit a frozen yoghurt store for a photo-op. Rather than criticize his suggestion, a colleague instead name calls him. He name calls her back.
Authority bias: The vice president picks up a piece of paper and pretends to be examining it while walking down a corridor, in an attempt to create the impression that she’s erudite in front of a passing visitor.
Euphemisms: Using palatable words to obscure unpleasant actions, a member of staff says, “That’s lying” to which the other objects, “Creative semantics.”
In-group bias*: Jonah’s dismissive comment toward Mike and Dan shows favoritism towards his White House group compared to the vice president’s staff—“Hey, it’s the flunkies.”
Moving the goalposts: Demanding different evidence when evidence for the original argument is provided. A member of staff points out that he’s the closest one to the vice president. Another member staff asks, “So would you take a bullet for her?” to which the first one rightfully responds, “That’s not my job.”
Self-serving bias*: “That is so great for me.” The vice president’s immediate reaction to a task force being green-lit is to think of herself rather than the country.
Episode 3 · Catherine, 4 examples
I spotted 1 example, GPT-4 spotted 4 examples (*), and we had one overlap (**).
Ad hominem*: The vice president dismisses a journalist’s criticism by attacking his physical appearance, not the content of his writing. “Stupid [expletive] [expletive].”
Confirmation bias*: “‘Cause l’ve met some people, okay? Real people, and I’ve got to tell you, a lot of them are … idiots.” The vice president dismisses public opinion, confirming her preexisting negative view of the general public’s intelligence.
False dilemma*: “Because the First Lady is getting one, so now I can't get one.” The Vice President suggests there are only two possible options. Either she gets a dog or the First Lady gets one. Both can’t get dogs.
It’s complicated**: The vice president exaggerates the complexity of a vain action she took, to dismiss her daughter’s criticisms that it was the wrong action.
Episode 4 · Chung, 8 examples
I spotted 4 examples, GPT-4 spotted 4 examples (*), and we had no overlaps.
Appeal to emotion: Governor Chung, having just been slighted by the Vice President for not being a real American makes a statement in which he says, “I’ve got a Purple Heart on my chest, but the one that beats inside of me is red, white, and blue.”
Appeal to fear: The vice president’s suggestion to get involved with immigration reform is met with resistance from her staff, “Once you go down that dark country road, the lights go out and the scarecrows come to life.”
Appeal to prejudice*: The vice president’s chief of staff’s statement about preserving "a certain kind of American identity", while disingenuous and cynical, caters to xenophobic attitudes for political leverage.
Dog whistling*: References to neighbors coming in to “crap a butt-full of drugs and beans on your flowers” and “certain kind of American identity” are xenophobic digs that aim to appeal to certain voter bases.
Plausible deniability: A senator who had previously promised joining the vice president’s filibuster reform runs into her team at an event. He then shares the elusive line, “l have the power to withdraw, which is what I’m gonna do right now,” leaving the staff unsure if he means from the conversation or from filibuster reform.
Red herring: Dan is blamed for a decision to put Sidney Purcell on a task force, to which Dan responds, “I blame George Washington. If he hadn’t started this whole form of government, then we wouldn't …”
Scapegoating*: In a conversation with Senator O’Brien, he blames “bad neighbors” for America’s problems.
Stereotyping*: The conversation about immigrants “taking over our ports" and the reference to the White House as “not the Yellow House” are steeped in racial stereotypes.
Episode 5 · Nicknames, 8 examples
I spotted 2 examples, GPT-4 spotted 7 examples (*), and we had one overlap (**).
Ad hominem*: The vice president’s personal attack on Jonah, referring to him as an “ape” and “useless waste of carbon,” are attacks on his character rather than refutations of things he has said.
Appeal to emotion. “You know, l am from oil, but that doesn’t mean that l don’t care deeply about my children’s future and, of course, their children’s future.”
Cognitive dissonance*: The vice president is frustrated with the president’s reprioritization of legislation, which includes him dropping her clean jobs initiative. She has to simultaneously accept the president’s judgment and hold disdain for his decisions.
Confirmation bias*: The team’s selective attention to certain information, like focusing on blog rumblings and nicknames, shows confirmation bias, where they pay attention to details that support their preconceived notions or fears.
False dilemma*: Does the vice president vote for the clean bills amendment she cares so much about and risk appearing disloyal to the president, or does she not do that and maintain her loyalty? No third option is considered.
Groupthink*: The vice president’s team often exhibits groupthink, where they align their opinions and strategies based on perceived group consensus or the dominant voice, rather than critically evaluating different perspectives.
Hyperbole, appeal to fear**: While playing the role of a critic, Dan’s portrayal of the clean jobs bill as “class genocide” is an appeal to fear, using exaggerated language to create a sense of threat among ordinary Americans.
Self-serving bias*: The self-serving bias includes attributing successes to ourselves and failures on external factors. The vice president’s frustration with the president’s actions shows a tendency to view herself as a victim of circumstances, avoiding responsibility for any shortcomings in her political strategy or relationships.
Episode 6 · Baseball, 5 examples
I spotted 3 examples, GPT-4 spotted 4 examples (*), and we had two overlaps (**).
Appeal to the bandwagon*: The group photo with the nutritionists and the Orioles players can be seen as an attempt to align with popular figures or causes to gain support or legitimacy.
False dilemma**: “So we’re looking at a wedding. Or a suicide. I haven’t decided which.” The vice president presents two extremes as the only options for handling her potential pregnancy, ignoring other possible solutions.
False equivalence: Dan asks Amy if the vice president is pregnant. Amy responds, “You think that as a woman I might sense some disturbance in the … lady matrix?” Amy is assumed to have some special insight into pregnancy tests on account of being a woman.
Misleading vividness*: “I’ve got to hang out with a bunch of boring grown-ups. Bleh.” The vice president uses this exaggerated description to create a stark contrast with the children’s environment, aiming to appear more relatable and likable to them.
Straw man**: “You don't find it repugnant that Selina Meyer
let a Secret Service agent go for smiling?”, “You’re calling the Vice President repugnant?” Mike twists a criticism about the vice president’s behavior into an exaggerated personal attack, deflecting from the actual issue.
Episode 7 · Full Disclosure, 7 examples
I spotted 2 examples, GPT-4 spotted 6 examples (*), and we had one overlap (**).
Ad hominem*: When Vic and Rod discuss Amy Brookheimer as “Blame Me Brookheimer” on the radio, they attack her character without addressing the actual issue. The goal is character assassination rather than engaging with facts.
Appeal to hypocrisy (tu quoque)*: Mike deflects criticism about his own professional mistakes by pointing out that others also have made mistakes, rather than addressing the criticism directly.
False cause*: The notion that firing a staff member will solve the office’s public relations crisis assumes a direct cause-and-effect relationship without considering other factors that contribute to the crisis.
False dilemma*: The vice president either fires someone on her staff to address the PR disaster she’s faced with, or she doesn’t and she continues to face it. No other options are presented.
Plausible deniability: A staff member points out that each of the things that is causing the vice president bad publicity can be spun in a way that shifts blame away from the vice president. For instance, they didn’t put two oil guys on a task force since they were “technically” ex-oil.
Red herring (obscurity through transparency)**: “[B]y showing that we have nothing to hide, then we can actually hide some stuff.” Much like ambiguity through precision, the vice president directs her team to release all their correspondence as a distraction tactic, diverting attention from specific issues by flooding the media with irrelevant information.
Scapegoating*: The vice president’s team considers blaming Sue for the issues in the office, trying to find an easy target to shift blame and responsibility away from themselves.
Episode 8 · Tears, 10 examples
I spotted 5 examples, GPT-4 spotted 7 examples (*), and we had two overlaps (**).
Ad hominem*: Congressman Furlong insults Dan by calling him names like “[expletive] nugget” and “buddy” in a derogatory manner. This is an attack on Dan’s character rather than addressing anything he has said or done.
Appeal to emotion**: The team’s strategy is to engineer the vice president crying on camera to gain public sympathy.
Appeal to the bandwagon*: The vice president is asked to endorse Congressman Furlong because it’s the trending thing to do.
Confirmation bias**: The vice president says, “How could that many Americans not like me? I think they got it wrong. They counted those numbers wrong or something. Or all those Americans are completely wrong.”
Euphemisms: The vice president announces a “realignment”, which some staffers understood to mean some will be promoted and some will lose their jobs.
False dilemma*: The idea that the vice president either endorses Congressman Furlong and faces political backlash or doesn’t endorse him and appears weak. Only two options are considered.
Guilt by association: A congressman asks the vice president not to publicly endorse him, on account of her being “damaged goods”.
Halo effect: At a dinner event, $50 donors don’t get the vice president’s time, but Platinum donors do. People who donated more are assumed to be worthy of her time.
Manipulation of information*: The decision to issue a press release about Dan’s promotion without considering the potential consequences shows a manipulation of information for personal gain.
Straw man*: Congressman Furlong constructs a straw man argument by grossly exaggerating the vice president’s political situation, comparing her endorsement to running over puppies with a campaign bus.
I hope that was enjoyable. Let me know if there are other good political satires you’d recommend I do a deep-dive of.
Until next time,
—Ali
Armando Iannucci has written some excellent shows and movies.
Really nice article. I personally enjoyed reading it and decided to use some the examples in my future classes.
Very interesting. So what we conclude is that AI machines have no sense of satire, irony, or humor. Perfect for our upcoming New Puritan society. Thank goodness for humans!