Prospect Theory

Why it matters

People don’t react to outcomes; they react to changes — gains and losses measured from wherever they happen to be standing — and that one shift bends almost every risky choice we make.

For example: a hospital faces a flu outbreak expected to kill 600 people and has to pick between two response plans. Told that Plan A “saves 200 for certain” while Plan B offers “a one-third chance of saving everyone and a two-thirds chance of saving no one,” most people grab the sure save. Tell a different room the same two plans as “400 die for certain” versus “a one-third chance no one dies and a two-thirds chance everyone dies,” and most people now take the gamble. The plans are identical — only the word saved became the word die — yet the choice flips. Nobody is weighing the outcomes. They’re weighing them as gains or as losses, against a reference point, on a curve that bends one way for each.

What it reveals. Whether a choice is tracking the actual outcomes or just the way those outcomes have been coded — as a gain to be protected or a loss to be escaped — relative to a reference point that someone, often invisibly, has set.
How it changes the read. You stop asking “which option is better?” and start asking “better against what baseline — and would the ranking survive if the same outcomes were described from the other side of it?”
When to foreground it. Any consequential choice carrying real risk or uncertainty, especially when the options can be told as a sure thing versus a gamble, or as a loss avoided versus a gain pursued.
What you’d miss without it. That people flip from cautious to reckless depending only on the framing — risk-averse when they see gains, risk-seeking when they see losses — and that tiny probabilities get wildly over-weighted, so the felt value of an option can swing without any change in its actual odds.
Where it misleads. Treated as a precise formula it overreaches — it gives reliable direction, not exact magnitudes. And it is a framing lever, so using it to push someone toward the option that merely sounds safest, rather than to surface a distortion in their own reasoning, is manipulation wearing the costume of analysis.

How to invoke it in Ora

You have a real, weighty decision in front of you — go or stay, build or buy, take the offer or hold out — and you want it structured properly, with the alternatives, the odds, the people affected, and what could go wrong all laid out at once.

Describe the decision and the options, and ask:

“Architect this decision: should we launch the new product now or wait a year? Walk the alternatives, the odds on each, what could go wrong, and whether we’re reading any option as a sure thing or a gamble just because of how it’s framed.”

Prospect theory is one of the always-loaded reasoning tools in the Decision Architecture analysis. As Ora lays out each alternative and weighs its probability-weighted outcomes, this lens checks how those outcomes are being valued — whether an option is winning or losing on its real merits, or only because it’s framed as protecting a gain or escaping a loss, and whether a long-shot probability is being treated as far larger than it is.

One thing to know: the words decision architecture, full decision analysis, big decision, should I do X or Y taking everything into account, or a full structured-decision request are what route you here. Saying prospect theory on its own does not summon the analysis — the model rides along inside it. The full analysis takes ten-plus minutes; if you only want a quick gut-check, a lighter decision pass is the better fit.

Name what you’re treating as your baseline — the status quo, what you’ve already spent, what you expected to happen. The whole gain-or-loss coding hangs on that reference point, and once it’s explicit you can ask whether it’s the right one or just the one you started from. If you can, say the odds out loud too: prospect theory’s second blade is probability weighting, and it bites hardest on the rare, vivid outcomes.

One thing Ora won’t do: weaponize the framing. It uses the gain/loss reading to surface a distortion in your own decision, tested against what you’d endorse on reflection — never to talk you into whichever option was dressed up to sound least like a loss.

How it works

Picture a public-health team staring down an outbreak. The projections say 600 people will die, and the team must choose between two response programs. The briefing puts it plainly: Program A will save 200 people for certain. Program B is a gamble — a one-in-three chance it saves all 600, and a two-in-three chance it saves no one. Faced with this, the overwhelming majority reach for the certain rescue. Two hundred lives, guaranteed, beats rolling the dice on everyone. It feels not just reasonable but obviously right.

Now the same outbreak, the same two programs, handed to a different team — only the briefing is worded the other way. Program C means 400 people will die for certain. Program D carries a one-in-three chance that nobody dies and a two-in-three chance that all 600 do. This time the majority refuse the certain option and take the gamble. A guaranteed 400 deaths is unbearable; better to bet on the slim chance of losing no one. That, too, feels obviously right.

Here is the catch. Program A and Program C are the same program — 200 saved is 400 lost, out of 600. Program B and Program D are the same gamble, number for number. Nothing changed but a single word: save became die. And yet the same kind of person, reasoning carefully, walked out cautious in the first room and reckless in the second. The economists Daniel Kahneman and Amos Tversky, who ran exactly this experiment, drew the only conclusion the result allows: people were never evaluating the outcomes at all. They were evaluating gains versus losses — and “200 saved” lands as a gain to be locked in, while “400 dead” lands as a loss to be fled.

What organizes the whole pattern is a reference point: the state of affairs people measure from. Above that line, an outcome reads as a gain; below it, as a loss. And the mind treats the two sides asymmetrically. On the gain side we turn cautious — a bird in the hand, the certain 200 over the risky 600. On the loss side we turn into gamblers — anything for a shot at making the loss disappear, the desperate bet over the certain 400 gone. Kahneman and Tversky drew this as a value function: an S-shaped curve, gently bending so that gains feel good with diminishing intensity, and steeply bending on the loss side so that losses bite far harder than the matching gain pleases. (That steeper drop for losses is the piece usually singled out and studied on its own as loss aversion — the part of the curve where a loss hurts about twice what an equal gain feels good.) Shift the reference point and the same outcome slides from the gain side to the loss side and back, which is exactly what the word swap did to the health programs.

There is a second strand, just as strange. We don’t take probabilities at face value either — we distort them. Tiny chances get blown up out of all proportion, while near-certainties get quietly shaved. This is why the very same person will buy a lottery ticket and an insurance policy in the same afternoon. The lottery’s microscopic odds of winning feel meaningfully larger than they are, so the long shot is worth a dollar; the microscopic odds of the house burning down feel large enough to dread, so the premium is worth paying. Neither is irrational in isolation. Together they reveal a mind that systematically overweights the rare and vivid. Put the two strands side by side — outcomes judged as gains and losses from a movable reference point, and probabilities bent by how they feel — and you have the engine Kahneman and Tversky called prospect theory: not a tidy account of how people should choose under risk, but a startlingly accurate map of how they actually do.

Framework & implementation

This section uses Ora’s own terms for the parts of an analysis, so that if you open the actual mode and lens files they line up. Each is glossed in plain language on first use.

Pipeline execution

Prospect theory is one of the always-loaded mental models in the Decision Architecture analysis — it sits in the mode’s ANALYTICAL PERSPECTIVES block under “always loaded,” available as a reasoning tool throughout the run. It is not the mode’s structure (Decision Architecture is a molecular mode that composes four full sub-analyses); it is the lens that governs how each alternative’s outcomes are valued. Where its partner lens loss aversion isolates one feature of the curve — the steeper slope for losses — prospect theory is the whole parent framework: the full value function (gains and losses scored from a reference point, not absolute states) and the probability weighting that distorts the odds. The mode runs at Gear 4, Ora’s most thorough setting — a Depth analyst and a Breadth analyst work the decision in parallel, critique each other, revise, and a consolidator integrates what survives.

Composition. Decision Architecture runs four components — decision-under-uncertainty (probability-weighted outcomes), constraint-mapping (binding constraints), stakeholder-mapping (who’s affected), and pre-mortem-action (failure pathways) — and fuses them through four synthesis stages: decision-frame-integration → stakeholder-impact-overlay → failure-mode-stress-test → integrated architecture. Prospect theory does its sharpest work at the front of that chain. It bites hardest on the output’s Alternatives with probability-weighted outcomes section — the place where each option’s odds-and-outcomes are scored — and it is the model the decision-frame-integration stage leans on when it sets the Decision frame, because how the frame is drawn fixes the reference point that decides which outcomes read as gains and which as losses.

Where the lens engages. It activates on its Detection Signals — the same option accepted under one framing and rejected under another; a decision-maker who is risk-averse on the upside but risk-seeking on the downside (or the reverse); small risks insured against while long-shot bets are taken at the same time; a reference point (the status quo, an expectation) visibly anchoring how each option is valued; behavior that diverges from straight expected-value. Its Application Steps run inside the analysis: identify the reference point the decision-maker is actually using; recognize that an outcome framed as a loss will pull harder than the same outcome framed as a gain; check for probability distortion — small risks over-weighted, near-certainties treated as sure things; and, when alternatives are being valued, test how shifting the reference point would re-rank them. This guards the integrity of the Alternatives with probability-weighted outcomes section: an alternative should not climb or sink in the ranking merely because of which side of an arbitrary baseline its outcomes were narrated from.

Cross-adversarial evaluation. At Gear 4 each analyst’s reading is critiqued by the other, which is where the lens’s own failure modes are caught, keyed to its Critical Questions: assuming the analyst’s reference point is the decision-maker’s (reference-point assumption); reading a precise percentage off a curve that only supports a qualitative direction (quantitative overreach); and the standing ethical check — has the gain/loss framing been chosen to inform the decision-maker, or to manipulate them? The evaluator presses each alternative: is this option genuinely stronger, or only stronger as told — and would the leading choice still lead if the frame were redrawn?

Integration and output. The consolidator carries the lens’s findings through the four synthesis stages into the integrated architecture. The Decision frame names the reference point it is working from rather than smuggling one in; the Alternatives with probability-weighted outcomes survive a reframing check; and the Recommended alternative with residual risks must not be an artifact of how the options were worded. The mode’s standing discipline is honesty about residual uncertainty — it does not produce clean recommendations that hide residual risk — and a recommendation that looks clean only because a loss was framed away is exactly that kind of hidden risk made visible.

What the analysis will not do. It will not use the framing to manipulate — the lens’s manipulation-framing failure mode is an explicit guardrail: the gain/loss reading is deployed to surface a distortion in the decision-maker’s own reasoning, tested against their reflective preferences, never to drive a choice they wouldn’t endorse. And it keeps the model qualitative where the evidence is qualitative, reporting the direction of a distortion rather than inventing a false precision the value function cannot bear.

Origin and evidence

Prospect theory is Daniel Kahneman and Amos Tversky’s, introduced in their 1979 Econometrica paper “Prospect Theory: An Analysis of Decision under Risk” — among the most-cited papers in the history of economics, and the work for which Kahneman received the 2002 Nobel in economics (Tversky having died in 1996, and the prize not awarded posthumously). The paper replaced the expected-utility account of choice under risk with two departures from it: a value function defined over gains and losses from a reference point rather than over absolute wealth — concave for gains (hence risk-averse on the upside), convex and steeper for losses (hence risk-seeking on the downside and loss-averse) — and a probability weighting function that overweights small probabilities and underweights large ones. The Asian-disease framing experiment recounted above is the canonical demonstration that the reference point, not the outcome, drives the choice. Tversky and Kahneman extended the theory in “Advances in Prospect Theory: Cumulative Representation of Uncertainty” (1992), whose cumulative form weights ranked outcomes rather than individual ones, fixing technical problems in the original and handling gambles with many outcomes. Kahneman’s Thinking, Fast and Slow (2011) is the accessible synthesis. The framing effects and the characteristic risk-attitude reversal are among the most replicated findings in behavioral economics, recovered across populations and domains, though the precise curvature and weighting parameters vary by context.

Applications and common uses

Prospect theory is a working tool wherever a choice involves risk, uncertainty, or gain/loss framing — used both to predict how real people will decide and to design the choices they face.

Pricing and marketing. Discounts framed as a loss avoided (“don’t miss out”) outperform the same discounts framed as a gain, and the over-weighting of small probabilities is the engine behind everything from extended warranties to sweepstakes.
Insurance and finance. The simultaneous appetite for insurance and lotteries — paying to erase a tiny risk while paying for a tiny chance — falls straight out of the weighting function, and the risk-seeking-in-losses pattern explains why investors double down on losing positions to claw back to even.
Negotiation. What each side treats as its reference point decides whether a given term reads as a concession (a loss) or a gain, so reframing the baseline can move a deal that looks deadlocked on the numbers alone.
Policy and public health. Whether a program is described in terms of lives saved or lives lost — or a tax in terms of what you keep versus what you forfeit — measurably shifts public support, which is why responsible communication names the framing rather than exploiting it.
Personal high-stakes decisions. Career moves, big purchases, and medical choices are routinely distorted by an inflated dread of the downside and an over-weighting of rare bad outcomes; writing each option’s odds and outcomes against an explicit baseline, then re-reading them from the other side, is the standard corrective.

In every case the move is the same: find the reference point, check whether the outcomes are being valued as gains or as losses because of it, watch for probabilities being bent by how they feel, and ask whether the choice would survive an honest reframing.

Failure modes and when not to use it

The lens’s characteristic ways of going wrong are catalogued in its Common Failure Modes:

Quantitative overreach. Treating the value function as a precise mathematical model and reading exact percentages off it. The tell is a prediction that specifies magnitudes the underlying parameters don’t support. Use the model for direction — which way the distortion runs — and calibrate magnitude empirically, never from the curve alone.
Reference-point assumption. Assuming the analyst’s reference point is the one the decision-maker is actually using. The tell is a prediction that fails in the field because the baseline was guessed wrong. Ask or measure the reference point directly rather than imposing the obvious one.
Manipulation framing. Turning the model into a manipulation toolkit — choosing the gain/loss framing to disadvantage one party without informing them. The tell is a set of framing choices that systematically tilt against the decision-maker’s interest. Hold to an informed-consent standard and disclose the framing; the same asymmetry that surfaces a distortion becomes exploitation the moment it’s turned against the actor.

When not to reach for it. When the choice involves no meaningful risk or gain/loss framing — outcomes are genuinely being judged in absolute terms with no reference point in play — the lens has nothing to grip, and importing it invents a distortion that isn’t there. When the decision turns on hard constraints or clean decision criteria rather than how outcomes are felt, a constraint-mapping or multi-criteria read is the better instrument. And where what looks like a prospect-theory effect is in fact ordinary, rational risk-aversion under the decision-maker’s real circumstances — a downside they genuinely cannot absorb — the asymmetry is a constraint to respect, not a bias to correct; for the narrow question of the loss-versus-gain steepness on its own, its partner lens loss aversion is the sharper tool.

Decision Architecture — the analysis this lens informs; integrates outcomes, constraints, stakeholders, and failure modes into one structured recommendation, valuing each alternative’s probability-weighted outcomes through this model.
Loss Aversion — the component lens: the steepness of prospect theory’s value function on the loss side, where a loss weighs about twice an equal gain. Prospect theory is the parent; loss aversion is one feature of its curve.
Framing Effect — the broader phenomenon that the same outcome, told as a gain or a loss, flips the choice; prospect theory is the mechanism that explains why the framing bites.
Anchoring — a sibling reference-point effect: an initial value silently sets the baseline that later judgments, including gain-or-loss reckonings, are measured against.