Dario's made-up 2-3x multiplier sparks debate over EA's quantified ethics

A fight over the methodology of quantified ethics has been brewing over the past few days. It started when researcher Sasha Gusev criticized Dario Amodei for pulling a "adult experience delivers a 2-3x improvement" multiplier out of thin air in his argument — while the EA (effective altruism) community spent the past week defending such practices as "rigorous and transparent." The core question: do numbers make fudging harder, or do they make it easier to disguise intuition as the "optimal solution"?

Confirmed

• Sasha Gusev pointed out the reverse problem also exists: mathematization lets EA practitioners pick a direction by gut feel first, then invent numbers to rationalize it, transforming intuition into a "morally optimal" choice
• David Manheim defended the approach: transparently estimating a number by intuition for reasoning is better than winging it entirely on vibes about what matters; Dario's critics haven't offered any alternative numbers — what they oppose is using math at all
• Manheim also pushed back on the claim that "numbers just rationalize intuition," arguing that being unable to use numbers is itself revealing
• quetzalrainbow took the defense's side: for any set of numbers you can construct "inconsistency" objections, but precisely because there are checkable numbers, faking becomes harder
• Itai Sher criticized certain circles (including EA) for settling into a mindset of "a made-up number is better than no number"; DrAtoosa responded that these circles treat "experience numbers" as inherently objectively good, obsessed with metrics generated poorly, measured messily, and rooted in inconsistent notions

Why it matters

The debate strikes at the role of quantification in AI governance and philanthropic decision-making: when "back-of-envelope numbers" are treated as rigorous analysis, they can either constrain intuitive bias or dress up predetermined conclusions in mathematical clothing. Dario's 2-3x multiplier has become the concrete flashpoint, and the discussion has moved beyond the specific number to the boundaries of quantitative reasoning in ethical decision-making.

2026-11-08 ~ 2026-11-09 · 6 related posts

• Mathematizing ethics lets EAs dress up vibes as optimal choices, researcher argues — SashaGusevPosts · 2026-11-08
• Numbers make things harder to fake, argue defenders of EA quantification — quetzal_rainbow · 2026-11-09
• [source] Researcher accuses Dario of inventing a 2-3x experience multiplier, igniting EA debate on numbers vs vibes — SashaGusevPosts · 2026-11-09
• [source] Itai Sher sparks debate: 'a bullshit number is better than no number' in AI circles — Dr_Atoosa · 2026-11-09
• [source] Making up numbers transparently beats hiding behind vibes, researcher argues in EA debate — davidmanheim · 2026-11-09
• Explicitly made-up numbers beat hidden vibes, argues AI safety researcher — davidmanheim · 2026-11-09