Herbert Simon originally proposed bounded rationality to account for deviations from fully rational decision-making arising from limited computational resources. An information-theoretic formulation measures processing costs as the relative entropy between a posterior decision strategy and a fixed prior strategy. In settings with multiple environments this formulation becomes equivalent to the rate-distortion problem of lossy compression, so that the optimal prior and posterior strategies are recovered by the Blahut-Arimoto algorithm. Because that algorithm relies on explicit partition sums, a sampling-based update rule was derived that adapts the prior strategy while converging to the same rate-distortion optimum. Complementary algorithmic work shows that heuristics extracted from relaxed puzzle instances accelerate Q-learning within answer-set-programmed Markov decision processes, reducing steps needed to reach optimal policies across deterministic, non-deterministic and non-stationary variants. Metaheuristic performance itself can be isolated by comparing outcome distributions against placebo counterparts, yielding three scalar measures—benefit, equivalence and risk—anchored to a pre-chosen threshold of practical significance. These results collectively illustrate how bounded agents replace exhaustive optimization with resource-aware heuristics whose quality can be quantified and improved.
Prospect theory as described in the supplied web research shows that decision makers evaluate outcomes relative to a reference point rather than final wealth and apply psychologically biased weights to gains losses and probabilities instead of computing expected value. The value function is concave over gains which produces risk aversion when people face certain gains and convex over losses which produces risk seeking when people face certain losses of comparable magnitude. Losses are weighted more heavily than equivalent gains a pattern labeled loss aversion. People also overweight small probabilities and underweight moderate to large probabilities so that rare events receive inflated decision weight. The same research notes that framing or editing of prospects can shift choices even when objective payoffs remain identical. These mechanisms are applied in the model of Klein and Deissenroth arXiv 1808.05572v1 which reproduces observed German residential solar photovoltaic adoption patterns by treating investment timing as sensitive to changes in profitability relative to the status quo rather than absolute profitability alone. The representation developed in Charles-Cadogan arXiv 1206.2665v1 formalizes the loss aversion index as an unobserved gauge transformation on Markowitz-Tversky-Kahneman reference points and treats the value function as the solution to a Dirichlet problem on the utility hypersurface. Together the sources establish that reference dependence loss aversion and nonlinear probability weighting jointly determine choices under risk.
Loss aversion and reference dependence appear in expected utility models that employ state-dependent linear utility functions for monetary outcomes, where the utility profile becomes steeper in loss regions relative to a reference level. Lahiri shows this formulation preserves characterizations of first-order stochastic dominance and mean-preserving spreads while linking increasing-concave profiles to risk aversion, then applies it to derive the partial-coverage contract a monopolist offers in an insurance market. Kleinberg, Kleinberg, and Oren embed the same principles in optimal stopping, letting the best observed option set a shifting reference point that adds an explicit utility penalty whenever a later selection falls below it; their analysis demonstrates how this bias raises reservation thresholds compared with risk-neutral benchmarks in sequential search. Classic experiments confirm the underlying patterns, with subjects displaying risk aversion for gains framed above a reference point and risk seeking for equivalent losses below it, and loss magnitudes weighted roughly twice as heavily as gains. Field patterns in asset markets reproduce the same asymmetry, as investors sell winners more readily than losers whose prices remain below purchase-price references, while homeowners facing nominal losses set higher list prices and delay sales. These mechanisms together illustrate how reference dependence alters both static contracts and dynamic choice without requiring nonlinear probability weighting.
Anchoring bias systematically pulls numerical judgments toward an initial value known as the anchor, with people failing to adjust sufficiently away from it even when the anchor is arbitrary or irrelevant. Individuals rely heavily on the first number encountered when making estimates under uncertainty, causing subsequent judgments to assimilate toward that starting point rather than an objective value. Classic experiments demonstrate that random numbers such as those generated by a wheel of fortune can bias estimates of probabilities or quantities, operating automatically outside conscious awareness. The effect appears across domains including general knowledge, visual judgments, math calculations, and probability assessments, producing lower estimates with low anchors and higher ones with high anchors. In negotiations the first offer serves as a powerful anchor that frames all subsequent counteroffers and concessions relative to it, often giving an advantage to the party making the initial proposal by influencing perceived fairness and final outcomes. These patterns hold despite incentives for accuracy or warnings and even among those with higher cognitive reflection. Parallel evidence from large language models shows that anchors shift entire output distributions in models such as Gemma-2B, Phi-2, and Llama-2-7B, with Shapley-value attribution confirming that the anchors drive reweighting of log-probabilities.
The availability heuristic leads people to assess the probability or frequency of events according to the ease with which relevant instances can be brought to mind, a mechanism first formalized by Tversky and Kahneman. When evaluating risks such as heart attacks, plane crashes, or product failures, individuals substitute mental availability for objective data, so events that are vivid, recent, emotionally charged, or widely publicized become judged as more likely than base rates justify. Media exposure and social-network occurrences heighten this retrieval fluency, causing dramatic but statistically rare hazards to dominate perception while chronic, low-salience risks such as hypertension or ordinary traffic accidents are discounted. Empirical patterns confirm that ease of recalling examples of failure directly elevates judged likelihood, whereas the same ease for successes lowers it. Affect amplifies the effect because emotionally laden events are more memorable and therefore more available, shaping both frequency estimates and assessments of severity. The resulting distortions produce systematic overestimation of dreaded, well-publicized threats and underestimation of familiar dangers, shifting public attention and policy priorities toward hazards whose objective magnitude is modest. Because availability can decouple sharply from actual frequencies, probability estimates formed under this heuristic routinely misalign with statistical evidence.
The supplied primary papers examine cross-correlations among 201 NSE stocks from 1996-2006 via the eigenvalue spectrum of the correlation matrix, showing most eigenvalues fall inside random-matrix bounds while the largest tracks market-wide moves and intermediate ones are few and close to the bulk, indicating weak sector identity. A second paper reports that tick and daily returns on NSE and BSE obey a power-law tail with exponent near 3, while volume and trade-count distributions differ from NYSE patterns and stocks remain highly correlated. Two further papers introduce perception-aware metrics for detecting topical bias in person-related query suggestions and the SoWinoBias test set for latent gender bias in coreference systems. The web descriptions list numbered citations without any URLs or arXiv identifiers attached, so no statement about overconfidence, overestimation, overplacement, overprecision, trading volume, risk-taking, or market timing can be traced to a source that produced the described result.
Confirmation bias appears consistently across classic experimental paradigms in hypothesis testing and impression formation. In Wason’s rule discovery task participants receive the triple 2-4-6 and must identify the hidden rule by proposing further triples; most generate confirming instances such as 10-12-14 rather than potentially falsifying ones such as 2-3-4. The same confirmatory search bias emerges in Wason’s four-card selection task, where participants testing a conditional rule preferentially turn over cards that can affirm rather than refute the rule. Parallel results occur in Bruner, Goodnow, and Austin’s 1956 concept-discovery studies, in which participants test hypothesized category features by selecting instances they expect to match rather than critical counterexamples. Costabile and Madon’s 2019 experiments extend these findings to social judgment: after participants formed dispositional inferences about a target, they showed superior memory for consistent information, interpreted ambiguous behaviors in line with the initial inference, and preferentially sought additional confirming details. These patterns demonstrate that once a belief is adopted, subsequent information search, interpretation, and retention favor evidence consistent with that belief across abstract reasoning and person-perception domains.
According to prospect theory, positive and negative frames alter decisions by moving outcomes into gain or loss domains, which reverses risk preferences so that positive frames promote risk aversion while negative frames encourage risk seeking. People evaluate outcomes relative to a reference point and therefore code them as gains or losses. Positive frames present outcomes as gains and negative frames present them as losses, even when the payoffs remain identical. The value function is concave for gains, producing risk aversion, and convex for losses, producing risk seeking, while losses exert greater weight because of loss aversion. In the Asian disease scenario, positive framing in terms of lives saved leads decision makers to select certain options, whereas negative framing in terms of lives lost reverses the choice pattern and favors risky options. This reflection effect occurs because negative frames generate a stronger pull toward avoiding losses than positive frames generate toward securing gains. The same underlying value function therefore produces opposite risk attitudes once the reference point shifts the coding of identical payoffs from the gain domain into the loss domain.
A nudge is a small change in the choice environment that steers people’s behavior in a predictable way without banning options or materially changing incentives. In Thaler and Sunstein’s framework it works by changing how choices are presented rather than by forcing a decision. The choice architect designs the context in which decisions are made, such as the order of options, defaults, framing, or placement, so that people are more likely to choose a particular option while still remaining free to choose otherwise. A nudge preserves freedom of choice, does not significantly change economic incentives, is easy and cheap to avoid, and influences behavior through the structure of the decision context, not coercion. Placing fruit at eye level in a cafeteria illustrates the mechanism because it raises the probability of selecting the healthier item without removing any alternatives; banning junk food would not qualify because it eliminates options rather than merely rearranging their presentation. Libertarian paternalism therefore employs such interventions to guide outcomes that align with individual welfare while explicitly retaining the right to opt out at no material cost. The same principles apply across repeated choice settings where small alterations in visibility, sequence, or default status reliably shift aggregate behavior without requiring education, price changes, or regulatory mandates.
Evidence from a controlled lab experiment in arXiv 2006.14868v3 shows status quo bias strongly dominates the decoy effect during single choices among two or three money lotteries, producing no detectable decoy influence while causing a measurable fraction of participants to switch from risk-averse to risk-seeking selections when a risky option is preset as default. Subjects in that study reported focusing on the maximum payoff under the default and on winning probability without it, a pattern consistent with reference-dependent evaluation. Complementary field data establish that defaults also shift real-world retirement behavior, with automatic enrollment raising plan participation 35 percentage points after three months and 25 points after two years, causing more than one quarter of workers to remain at a 3 percent contribution rate despite available matching up to 6 percent, and leading one-third of enrollees to allocate all assets to the default fund. Broader adoption of such defaults routinely lifts participation above 90 percent. A systematic review in arXiv 2212.03283v1 links the same bias to user resistance during information-system implementations and identifies psychological and contextual factors that can reduce it.
Install this pack and your MIND begins smart — then every answer is grounded in your own knowledge graph.
Try MIND free →