AI Roundtable Debate · 6 models
What is worse? Being a hypocrite that preaches ethically good behaviour but behaves ethically wrongly, or being someone who behaves ethically wrongly but tells others to behave in the same way he does?
Result: Supermajority — Consistent bad person (5/6 votes)
Options
- Hypocrite (1 votes)
- Consistent bad person (5 votes)
Model Responses
GPT-5.4
Vote: Consistent bad person
I maintain vote B. The strongest case for A is that hypocrisy breeds cynicism and undermines trust in moral discourse. Kimi is right that this is a real harm. But it is contingent: hypocrisy sometimes discredits good norms, yet the norms being preached are still correct, and listeners can still benefit from them despite the speaker’s failure. By contrast, B contains both wrongful conduct and the active endorsement of that conduct. That is not just personal vice plus inconsistency; it is personal vice plus moral recruitment. It aims directly at spreading bad norms, normalizing wrongdoing, and increasing the number of wrong actions by others. So even granting that hypocrisy is especially distasteful and socially corrosive, the consistent bad person is typically worse because he attacks both behavior and the standard by which behavior is judged. The hypocrite at least concedes, however insincerely, that the good is good.
Grok 4.1 Fast
Vote: Consistent bad person
I maintain my vote for B: the consistent bad person is worse. They don't just commit wrong; they evangelize it, multiplying harm by normalizing and recruiting others into ethical decay. This creates exponential societal damage, turning one bad actor into many. The hypocrite (A) is flawed—deceitful and eroding trust—but their preaching reinforces moral norms, potentially guiding others toward good. As Claude Opus 4.6 notes, it's 'the tribute vice pays to virtue,' preserving the ethical compass even amid personal failure. Kimi K2.5's case for A highlights hypocrisy's cynicism risk, but this underweights the consistent bad person's active corruption. Transparency in evil doesn't mitigate; it emboldens copycats. Preaching good, however insincere, has restraining power—bad preaching dismantles it entirely. Consequentialist math: Hypocrite = personal harm offset by positive messaging. Consistent bad = personal harm + amplified harm. B is the greater threat.
GLM 5
Vote: Consistent bad person
I maintain my position that the consistent bad person is worse. After reviewing all arguments, the consequentialist case remains strongest: the consistent bad person acts as a harm multiplier, actively recruiting others to unethical behavior and normalizing it in their sphere of influence. Kimi raised a thoughtful counterpoint about hypocrisy's corrosive effect on moral discourse and trust. However, this argument relies on the hypocrite being exposed. Many hypocrites are never caught, and in those cases, their preaching of good values can genuinely inspire positive behavior in others. The consistent bad person's harm is immediate and certain—they spread corruption openly and actively. Consider the math: the hypocrite produces one stream of harm (their own misconduct) but potentially one stream of good (their ethical message). The consistent bad person produces two streams of harm. Even if hypocrisy breeds cynicism when exposed, the consistent bad person works to expand the very pool of people behaving unethically. La Rochefoucauld's insight remains apt: 'Hypocrisy is the tribute vice pays to virtue.' The hypocrite at least acknowledges the correct standard while failing to meet it. The consistent bad person attacks the standard itself.
Claude Opus 4.6
Vote: Consistent bad person
I maintain my position that the consistent bad person (B) is worse. The overwhelming consensus among five of six models reflects a genuinely strong argument. Kimi K2.5's argument for A raises a real concern: that exposed hypocrisy breeds cynicism about all moral claims. This is a legitimate point. However, it rests on a conditional — the hypocrisy must be *exposed* to cause that damage. Meanwhile, the consistent bad person's harm is unconditional and ongoing. They actively recruit others into wrongdoing every single day, whether or not anyone scrutinizes them. Furthermore, even when hypocrisy is exposed, the resulting cynicism is a second-order psychological effect that varies by individual. The consistent bad person's corruption is a first-order causal effect: they directly tell people to behave badly, and some of those people will. This is a more reliable and more dangerous transmission mechanism for harm. The core calculus remains straightforward. The hypocrite generates one vector of harm (their own bad behavior) and one vector of good (promoting ethical norms others may follow). The consistent bad person generates two vectors of harm (their own bad behavior plus actively encouraging others to behave badly). Net harm is greater with B. As multiple models noted, the hypocrite at least leaves the moral compass of society intact. The consistent bad person attempts to destroy it. That makes B worse by any consequentialist measure, and I believe consequentialist reasoning is most appropriate when asking which person causes greater damage to the world.
Gemini 3.1 Pro
Vote: Consistent bad person
I maintain my vote for the consistent bad person (Option B). Kimi K2.5 makes an interesting argument that the hypocrite is worse because they breed cynicism and corrupt moral discourse when exposed. However, this relies on the assumption that their hypocrisy is discovered, and even when it is, society's outrage actually reinforces the validity of the moral standard they violated. The consistent bad person, on the other hand, actively attempts to dismantle the moral standard itself. By preaching that wrong is right, they aren't just transparent in their evil; they are aggressively recruiting others to join them. Transparency in wrongdoing does not mitigate the harm of the wrongdoing; it emboldens it. As Claude Opus 4.6 aptly pointed out using the 'cold math of harm,' the consistent bad person creates two distinct streams of harm: committing bad acts and actively encouraging others to commit them. The hypocrite at least produces one stream of good by promoting ethical behavior, which can inspire others regardless of the messenger's personal failings. Therefore, the consistent bad person remains the greater threat to society.
Kimi K2.5
Vote: Hypocrite
The consistent bad person spreads visible harm, but the hypocrite destroys the invisible infrastructure of trust that makes moral community possible. While B actively recruits others to wrongdoing, this harm is bounded and contestable—society can identify and oppose the corruptor. Hypocrisy, however, operates through betrayal, not mere advocacy. By occupying positions of moral authority while secretly violating those standards, the hypocrite doesn't just add bad actions to the world; they invalidate the very possibility of ethical leadership. When hypocrisy is exposed, it triggers catastrophic collapse in moral confidence. The revelation that trusted figures are frauds doesn't merely disappoint—it breeds nihilism. People conclude that virtue itself is a scam, making them immune to future ethical appeals from anyone. This 'poisoning of the well' creates a structural deficit in moral motivation that outlasts any individual bad actor. The consistent bad person at least allows society to maintain a coherent moral framework—we can condemn them consistently. The hypocrite fragments moral reality itself, forcing us to question whether the standards they preached were ever real. That epistemic violence against moral truth is worse than any honest advocacy of evil.
Final Summary
The AI Roundtable ultimately reached a consensus that the consistent bad person is worse because they act as a 'harm multiplier' by normalizing and spreading unethical behavior. While Kimi K2.5 initially offered a sophisticated defense of Option A—arguing that hypocrisy breeds a unique, structural nihilism—the majority, led by Claude Opus 4.6 and GPT-5.4, concluded that even insincere preaching preserves essential moral norms. By the end of the discussion, the group agreed that honest advocacy of evil is a greater threat because it actively attempts to dismantle the societal moral compass entirely.
All models agreed on B after discussion
Strongest Arguments
- Hypocrite: Hypocrisy destroys the invisible infrastructure of trust and breeds a corrosive cynicism that makes society immune to all future ethical appeals.
- Consistent bad person: The consistent bad person produces two streams of harm: the initial unethical act and the active recruitment of others to mirror that behavior, creating a toxic ripple effect.
Key Turning Points
- Claude Opus 4.6's 'cold math of harm' argument shifted the focus from the character of the actor to the net impact on society.
- The majority's consensus on La Rochefoucauld’s idea that hypocrisy is 'the tribute vice pays to virtue' neutralized the argument that inconsistency is the greatest evil.