A Modern Guide to Thinking, Fast and Slow
Part V - Two Selves
Chapter 35: Two Selves
Overview
“Utility” has two distinct meanings: experienced utility (pleasure and pain as they unfold) and decision utility (the anticipated pleasure or pain we’ll get from a choice). There is a systematic mismatch between the two. Memory does not record experiences by summing moments over time. Retrospective judgments overweight the worst (or best) moment and the ending while largely ignoring how long the episode lasted.
This creates a conflict between two selves: the experiencing self, which lives the moments, and the remembering self, which evaluates them and then governs future decisions. Because choice is driven more by remembered episodes than by duration-weighted experience, people can end up optimizing for better memories rather than better lived experience and sometimes choose an objectively worse experience because it will be remembered as less bad.
Replications & Reliability
The peak-end rule and duration neglect in retrospective pain evaluations are well supported in short, well-defined episodes. The book’s core evidence comes from the colonoscopy study (Redelmeier & Kahneman, 1996) and a follow-up randomized trial that manipulated the ending of colonoscopy (Redelmeier, Katz, & Kahneman, 2003). A large meta-analysis of 174 effect sizes (Alaybek et al., 2022) found that the peak-end rule was robust across boundary conditions and that the effect of the duration was essentially nil, supporting the idea of duration neglect.
However, the peak-end rule appears less predictive for extended, multi-episode experiences (e.g., remembering vacations or “the previous day”), where other summaries can perform as well or better (Kemp et al., 2008; Miron-Shatz, 2009).
Recommendation
This chapter’s conclusions are best applied to short, well-defined episodes, where the peak-end rule and duration neglect are best supported. Be cautious about generalizing to extended, multi-episode experiences, where peak-end is often a weaker predictor and other summaries can matter as much or more.
Chapter 36: Life as a Story
Overview
We often judge experiences (and even whole lives) less by how they felt moment to moment and more by the story we can tell about them afterward. The remembering self turns life into a narrative built from peaks and endings, while largely ignoring duration. That’s why the last minutes of an opera, a final conversation, or a late reconciliation can feel enormously important, even if they add little time or change little of the lived experience.
This “life as a story” mindset shapes what we value and what we choose. We plan trips, take photos, and chase “memorable” moments because we are investing in the account our memory will keep, not just in present pleasure. In the extreme, we can become surprisingly indifferent to what our experiencing self endures if the episode won’t be remembered.
Replications & Reliability
- The “Jen” study (Diener et al., 2001): This study found that in judgments of whole lives, people heavily weight the ending and largely neglect duration. Adding mildly pleasant years to a wonderful life can make it seem worse, whereas adding moderately bad years to a terrible life can make it seem better, especially when the extra years are at the end. A 2025 direct replication found that the effect did not reproduce when ending information could not be contrasted, but did reproduce when participants could compare endings (i.e., when ending information was made salient by contrast) (Fujishima, 2025).
- The Spring break study (Diener et al., 2003): I am not aware of any replications. Please email me if you know of any.
Recommendation
The direct replication of the “Jen” finding strengthens the case that people can evaluate whole lives in a narrative, end-weighted way. The caveat is that the effect appears to depend on contrast—it is most reliable when the ending can be meaningfully compared to what came before, and weaker or absent when that comparison isn’t available. Overall, treat the chapter as broadly reliable on the direction of the bias, but don’t assume the strongest “adding extra years that are not as good makes a life seem worse” effect will show up in every context—its presence depends on whether the ending is salient and comparable.
Chapter 37: Experienced Well-Being
Overview
In this chapter, Kahneman distinguishes between life satisfaction, which is the overall verdict of the remembering self, and experienced well-being, which is how life actually feels as it unfolds. A global judgment condenses everything into a story, but lived happiness is made of minutes (time spent in states you’d rather continue, time spent in situations you want to escape, and time spent in the neutral middle where you wouldn’t care either way). This distinction affects what “improving happiness” can mean.
Replications & Reliability
In this chapter, Kahneman claims that life evaluation rose steadily with income while their yesterday affect measures showed no further gains beyond ~$75,000/year (in that period’s dollars). This claim is tied to Kahneman & Deaton’s 2008–2009 analysis of ~450,000 U.S. respondents from the Gallup-Healthways Well-Being Index (Kahneman & Deaton, 2010). This is a controversial claim.
- A large experience-sampling study by Killingsworth that drew on over 1.7 million reports challenges the idea of a plateau. It found “no evidence for an experienced well-being plateau above $75,000/y” and argued that higher incomes are associated with both feeling better day-to-day and being more satisfied with life overall (Killingsworth, 2021).
- In a 2023 adversarial collaboration between Killingsworth, Kahneman, and Mellers (KKM), the authors argue that the discrepancy was due to using different measures and the findings can be reconciled. Kahneman & Deaton’s Gallup-style measures hit a ceiling effect at higher incomes, so those measures were better at detecting reductions in misery. When methods that look across the whole distribution of happiness were used, they found that for most people, experienced well-being keeps rising with income, but for the least happy minority, the gains flatten above roughly an inflation-adjusted threshold (around ~$100K).
- This was critiqued by Rohrer and Wenz (2024), who argued that KKM’s framing invites causal interpretations that the evidence does not warrant.
- In a published reply, KKM clarify that their goal was to reconcile associations, not establish causality: they explicitly describe their evidence as correlational, as were the earlier studies they were reconciling. They write that "in the handful of studies that can draw causal inferences between income and happiness, evidence suggests that having more money does, on average, lead to greater happiness" (Killingsworth, Kahneman, and Mellers, 2024).
Recommendation
This chapter is broadly reliable on the core distinction between life evaluation (a reflective judgment) and experienced well-being (how life feels moment to moment). Be more skeptical about the specific income claims—especially the idea of a “satiation point.” The original plateau result relied on measures that likely hit ceiling effects at higher incomes, making additional gains harder to detect, and the data are observational, so the findings are associations rather than causal evidence that income stops causing improvements past a cutoff. A safer takeaway is that income has diminishing returns for day-to-day feelings and the pattern varies across people.
Chapter 38: Thinking About Life
Overview
When people answer the question “How satisfied are you with your life?”, they often substitute an easier one: How do I feel right now? What’s most salient lately? What big change is on my mind? Because of this, small but recent events (like finding a coin) or highly available facts (like a recent marriage or an accident) can affect life satisfaction reports more than they change how life feels moment to moment. This is the focusing illusion: whatever you’re thinking about influences your judgment more than it should. The result is predictable affective forecasting errors and “miswanting” (overvaluing changes that will soon fade into the background as attention moves on).
Replications & Reliability
- The “coin on the copying machine” study is a famous demonstration of the “mood heuristic,” in which a tiny positive event (finding a dime) appears to raise people’s global life satisfaction. The original study appeared in Schwarz’s doctoral thesis (in German, “Mood as Information: On the Impact of Moods on the Evaluation of One’s Life,” Heidelberg: Springer Verlag, 1987), and it is described in later publications (e.g., Schwarz and Strack, 1999). The study had an extremely small sample (N=16), and in later reviews of the same work the difference is described as not statistically significant. A large paper that ran nine direct and conceptual replications with much larger samples found that most effects were non-significant, and when effects did appear they were far smaller than the early literature suggested (Yap et al., 2018).
- The focusing illusion in general: Modern research supports the underlying mechanism (often studied as focalism in affective forecasting) that people tend to overestimate how intense and lasting their feelings will be about a single future change because attention is narrowly centered on that change while the rest of life is underweighted. Go here for an overview of research in the area.
Recommendation
The idea that life satisfaction judgments are constructed—and therefore sensitive to what’s salient in the moment—is well-supported. But the famous “dime on the copier” study is a good reminder that fascinating, oft-repeated studies are not necessarily reliable. Before repeating a finding, it’s worth checking basics like sample size, whether the reported difference was statistically significant, and whether later studies (especially high-powered replications) reproduced the effect.
