{"id":"b1f8bd11-8e57-4e4d-ae86-162d8c86a8d9","arxiv_id":"1908.01709","paper_version":3,"verdict":"REJECT","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"high","formal_verification":"none","parameter_count":2,"one_line_summary":"Using three extreme-outcome gambling problems, the paper finds that most people prefer the certain option and argues that time-averaged wealth growth, not expected value, drives these choices.","lead":"This paper asks why people avoid extreme gambles even when the average outcome is attractive, and tests whether a best long-run wealth growth rule explains their choices. It reports three survey questions where most people pick the safe option, and proposes a contrast-ratio measure to say when risk-seeking disappears.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Central claim rests on an untested attribution: extreme-outcome choices are read as evidence of repetition-based time averaging, but no experimental condition isolates that mechanism from one-shot loss aversion or scale effects.","rationale":"The reader's weakest assumption is the same as the single most load-bearing concern: the experimental data cannot identify the mental process behind the choices. My own reading of Sections 3 and 4 confirms that the theory's contrast-ratio and S-curve are not derived from a direct time-average maximization rule with independently estimated parameters; the loss branch inserts rho and the fuzziness threshold is asserted. Thus the strong conclusion in Section 5 is not supported by the evidence. The proposed experiment would directly test the repetition attribution and would settle whether the central claim survives. Since the concern is unresolved, the REJECT verdict stands.","tokens_in":7323,"tokens_out":8370,"duration_ms":86476,"concrete_test":"Run a pre-registered online experiment (N>=300) presenting the same three problems in four between-subject framings: (1) original wording; (2) explicit one-shot wording ('you will face this gamble only once'); (3) explicit repeated wording ('imagine this gamble is played 100 times'); and (4) moderate-stakes equivalents (e.g., lose $50 vs 50% chance of losing $100; win $100 vs 5% chance of $2,000). If proportions in (1) and (2) are statistically indistinguishable and differ from (3), the indefinite-repetition inference is refuted. If (4) replicates the standard Prospect Theory pattern, extreme stakes - not time averaging - drive the choices.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section 5 claims 'strong evidence that decision-makers assume indefinite repetitions.' The only support is the three extreme-outcome choices in Section 4. In Problem 1, 98.5% choose the sure loss of half over a 50% chance of losing everything; in Problem 2, 95.5% choose the sure $10M over a 5% chance of $200M. The paper interprets these as 'consistent with the best time average' and cites a respondent comment as evidence that individuals 'see gambles as dynamic processes.' However, the questionnaire never mentions repetition, and no item probes whether respondents imagined repeated plays. The design does not rule out one-shot loss aversion, ruin aversion, minimum-wealth constraints, or simple non-linear utility of extreme amounts. The model in Section 3 is also partly post hoc: the loss-side rate rho is inserted to create small-loss risk seeking, and the +/-0.5 dB fuzziness threshold is asserted. If the repetition attribution is false, the data do not test the time-average model at all.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes that decision-makers evaluate risky prospects by implicitly assuming indefinite repetition and comparing time-average growth rates rather than one-shot expected values. Using rhetorical devices it labels 'meiosis' for gains and 'hyperbole' for losses, the author constructs an S-shaped value function from nonextensive-statistics functions, introduces a decibel-scale 'contrast ratio' to define fuzzy and crisp regions of choice, and reports three survey problems with extreme outcomes that are claimed to show respondents choosing options with the best time average. The central conclusion is that the data provide 'strong evidence that decision-makers assume indefinite repetitions' and that time should therefore be treated as part of the physical stimulus driving choice.","tokens_in":7496,"tokens_out":5770,"duration_ms":57803,"significance":"The paper engages a real and timely question: whether time-average (non-ergodic) reasoning, as developed by Peters and Gell-Mann, can explain deviations from Prospect Theory for extreme outcomes. The gain-side inequality in Section 2.1, showing px >= (1+x)^p - 1, is a clean and correct mathematical observation, and the idea of testing the time-average model with extreme-outcome problems is falsifiable. However, the theoretical model is only partially derived: the loss-side S-curve depends on an ad hoc rate rho, the fuzziness threshold is postulated rather than derived, and the experimental section reports simple percentages without statistical inference or a control for one-shot motivations. If the model and experiment were properly developed, the paper could be a useful contribution, but as it stands the evidence does not support the strong conclusion drawn.","major_comments":[{"comment":"The loss-side derivation is internally inconsistent. In Section 2.2 the hyperbolic curves are given as (1+px)^(1/p)-1 and stated to lie above the line x for -1 <= x < 0, which would make the sure-loss option l3 preferable; however, Eq. (5) defines the choice as max{e^{\\rho x}_p - 1, x} and assigns 'x for small losses' and 'e^{\\rho x}_p - 1 for big losses,' which reverses the ordering for rho = 1. The role of x is also ambiguous because l3 is defined as a loss of Mp, which has expected change px rather than x. This ambiguity undermines the derivation of the S-curve's loss side and must be resolved before the model can be evaluated.","section":"Section 3, Eq. (5) and Figure 2"},{"comment":"The rate rho is introduced specifically to reproduce the known pattern of risk seeking for small losses: the text states 'we must insert a rate rho into the hyperbolic argumentation process' without any independent derivation or empirical justification. Because the shape of the loss-side S-curve, and hence the predicted transition from risk seeking to ruin aversion, is controlled by this free parameter, the central theoretical claim is not a derivation from time-average dynamics but a post-hoc calibration to Prospect Theory's known pattern. The manuscript would need to derive rho from a stated principle or identify it independently from data.","section":"Section 3, paragraph after Eq. (5)"},{"comment":"The experimental design does not support the conclusion that respondents used implicit indefinite-repetition reasoning. The questionnaire never mentions repetition or time averaging, and no item asks respondents whether they imagined repeated plays; the only evidence offered is one quoted respondent comment. The observed choices could equally arise from one-shot loss aversion, ruin aversion, minimum-wealth constraints, or diminishing marginal utility of extreme amounts. Moreover, the paper reports only percentages for the three problems, with no confidence intervals, hypothesis tests, or baseline comparison to non-extreme versions of the same problems; the Shannon-entropy calculations in Section 4 are descriptive and do not establish statistical significance.","section":"Section 4, Problems 1-3"},{"comment":"The fuzziness threshold of -0.5 dB to 0.5 dB is asserted without derivation or independent measurement. The claim that this threshold 'can define a threshold between the stimuli and the sensations (or perceptions) they produce' is therefore not tested by the data. Since the paper's account of when risk seeking appears and disappears depends on this threshold, the central claim about fuzzy versus crisp regions is unsupported.","section":"Section 3, Fig. 3 and surrounding text"},{"comment":"The conclusion that the paper provides 'strong evidence that decision-makers assume indefinite repetitions' is not warranted by the reported results. Three extreme-outcome choices, without controls for rival mechanisms and without statistical testing, cannot bear the weight of this conclusion. The phrasing overstates what the data show and should be tempered substantially even if the earlier issues are addressed.","section":"Section 5, Conclusion"}],"minor_comments":[{"comment":"The title on the first page is 'The Time Importance for Prospect Theory,' while the arXiv metadata title is 'Behavioral Biases and Nonadditive Dynamics in Risk Taking: An Experimental Investigation'; this inconsistency should be fixed.","section":"Title and abstract"},{"comment":"There is a typographical error 'Kahnemam' instead of 'Kahneman' in the discussion of Problem 3.","section":"Section 4, Problem 3"},{"comment":"The notation in Figure 2 and the surrounding text is inconsistent: the text refers to curves (1+px)^(1/p)-1, the figure caption says '(1 + px)^{1/p}', and the expression exp is defined but not used consistently. Please harmonize the notation throughout.","section":"Section 2.2 and Figure 2"},{"comment":"The sample is described as psychology students from a single institution, but no demographic or recruitment details are provided, and no ethics approval statement is included.","section":"Section 4, Problems 1-2"}],"recommendation":"reject","confidential_remarks":"The manuscript reads as an early-stage working paper. The central empirical claim is not supported by the experimental design, and the theoretical model has at least one free parameter that is introduced purely to reproduce known results. A revision that adds a proper experiment with repetition framing, statistical tests, and a derived threshold would be needed to make the claims defensible; this is beyond the scope of a normal revision. The paper also relies heavily on the author's related unpublished arXiv manuscripts, which limits independent verification."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this is a speculative paper with one genuinely novel idea and an experiment too weak to carry the load. The novel piece is building a prospect-theory S-curve from time-average growth (Peters) plus a dB contrast ratio to define a \"fuzzy\" region; that framing might interest people working on non-ergodic decision theory. The three extreme-outcome problems are real survey data—67 students, but the percentages are striking and the problems themselves are a useful addition to the literature.\n\nWhat it does well: the gain-side algebra is a clean observation: concavity of (1+x)^p makes the time-average curve sit below the tangent px, so a repetition-based argument yields risk aversion for gains. The contrast-ratio dB idea is a plausible way to talk about discriminability of gambles. The author is transparent that the fuzziness threshold is hypothetical.\n\nThe soft spots are substantial. The loss-side S-curve depends on a rate ρ inserted specifically to reproduce the known small-loss risk-seeking pattern; no independent justification or estimation for ρ is given. The fuzziness threshold ±0.5 dB is asserted. The experiment has no controls, no statistical tests, and no baseline; the conclusions rest on three binary choices from a convenience sample. More importantly, the central attribution—that respondents are implicitly reasoning about indefinite repetition—is not tested. The survey never mentions repetition; the extreme stakes could just as well trigger one-shot ruin aversion, loss aversion, or non-linear utility of very large amounts. A participant's comment \"it is easier to continue with half than to start over\" is about wealth levels, not repetition. And the math has an internal inconsistency: Figure 2 and the text around equation (5) don't agree about the low-distinguishability region for p=0.5, and equation (5)'s piecewise form appears reversed for small losses.\n\nIf this lands on a desk, it's not a publishable research claim in its current form. But the underlying question—whether time-averaging heuristics explain why extreme outcomes flip prospect theory's predictions—is worth a serious look. I'd send it to a referee rather than desk-reject, because the model is novel and the data points, however crude, are real and potentially useful. The referee should push for a redesigned experiment that directly tests the repetition mechanism and for a principled derivation of ρ and the threshold.","headline":"A novel time-average S-curve with an interesting extreme-outcome experiment, but the central claim about repetition is unsupported and the model has ad hoc parameters.","tokens_in":8012,"tokens_out":5264,"would_cite":false,"duration_ms":49409,"reading_group":"maybe","serious_thinker":"no","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper argues that Prospect Theory's risk predictions flip for extreme outcomes because people evaluate gambles as repeated time-average growth choices.","keywords":["Prospect Theory","time average","contrast ratio","nonextensive statistics","risk taking","behavioral biases","extreme outcomes","ruin aversion"],"falsifier":"Present the same three problems to two groups: one told the gamble will occur exactly once, the other told it will repeat many times. If the certainty-choice rates are equally high in both groups, the paper's time-average mechanism fails; if the repeated frame raises certainty choices above the one-shot frame, the central claim is supported.","tokens_in":7086,"feed_emoji":"🎲","tokens_out":14639,"duration_ms":113976,"temperature":0.7,"pith_summary":"This paper argues that Prospect Theory leaves out a fundamental variable: time. When people choose between risky options, the author claims, they implicitly assume the gamble will be repeated many times and therefore compare the time-average growth of their wealth rather than a one-shot expected value. Modeling the S-shaped value curve with the deformed functions of nonextensive statistics, the paper predicts that extreme outcomes should eliminate the risk-seeking behaviors that Prospect Theory expects for small probabilities and for losses. An experiment with 67 students on three extreme-outcome problems supports this: the great majority chose the certain option in each case. If correct, the paper supplies a physical-stimulus account of when Prospect Theory's pattern flips and why.","feed_headline":"Extreme outcomes flip Prospect Theory's risk predictions","feed_subtitle":"A 67-person experiment finds people choose sure bets over long-shot riches and total ruin, favoring time-average growth.","key_machinery":"The central objects are the time-average growth functions $p\\ln_p(1+x)$ for gains and $\\exp_p(\\rho x)-1$ for losses, where $\\ln_p(x)\\equiv (x^p-1)/p$ and $\\exp_p(x)\\equiv (1+px)^{1/p}$ are the deformed logarithm and exponential of nonextensive statistics. The argument rests on the tangent relation between the linear expected-change option and these concave or convex curves at $x=0$, and on the contrast ratio between the two time-average signals, expressed in decibels, as the threshold separating fuzzy from crisp decision regions.","core_discovery":"The paper's central claim is that decision-makers facing known probabilities and outcomes assume indefinite repetitions, so the proper measure of a prospect is its time-average wealth growth rather than its ensemble-average expected value. For gains, the certain option of winning a fraction $x$ is compared with the risky option of winning the same expected amount with probability $p$; the author shows the time-average growth of the risky option is $p\\ln_p(1+x)$, which is always below the proportional gain $px$, yielding risk aversion. For losses, a hyperbolic construction gives $\\exp_p(\\rho x)-1$, which produces risk seeking for small losses but flips to ruin aversion when the potential loss is extreme. The contrast ratio between the two time-average signals, measured in decibels, defines a fuzziness region in which choices are unclear and a crisp region in which the better time-average dominates. The paper's three experimental problems, with extreme outcomes, find between 73% and 98% of 67 respondents choosing the certain option, which the author interprets as direct evidence for this time-average mechanism.","pith_inferences":["If repetitive time-average reasoning drives extreme-stakes choices, then explicitly framing the same gamble as a single one-shot event should substantially lower the certainty-preference rate; this is a direct test the paper does not perform.","The fuzzy boundary of about $\\pm 0.5$ dB is chosen by inspection rather than derived; a systematic experiment varying outcome magnitudes and probabilities could map decibel distance to choice-frequency curves and turn the fuzzy/crisp distinction into a quantitative psychophysical law.","The model suggests that the probability-weighting distortions in Prospect Theory are emergent from time averaging rather than a separate cognitive bias, which would imply that option framing (repeated versus one-shot) could shift risk preferences in real financial decisions."],"forward_implications":["At extreme gains and losses, the classic risk-seeking patterns of Prospect Theory flip: people choose the certain moderate outcome over a tiny chance of a huge gain and over a small chance of total ruin.","The S-shaped value curve of Prospect Theory can be rederived from time-average dynamics and nonextensive statistics, with the parameter $\\rho$ controlling the small-loss risk-seeking region.","The contrast ratio between time averages predicts decision crispness: low-contrast gambles produce fuzzy regions with less predictable choices, as the divided responses in one problem show.","Losses severe enough to threaten all of a person's possessions trigger 'ruin aversion,' which the paper treats as the time-average mechanism overriding the usual risk seeking for losses."],"supporting_citations":[{"why":"Provides the original Prospect Theory formulation and the experimental paradigm whose risk-seeking patterns this paper challenges.","marker":"[2]"},{"why":"States that risk seeking holds only when outcomes are not extreme, the exact gap this paper targets.","marker":"[8]"},{"why":"Supplies the time-average growth-rate analysis for repeated gambles that the model adopts.","marker":"[3]"},{"why":"Frames maximizing long-term time-average growth as rational, grounding the normative interpretation of the choices.","marker":"[4]"},{"why":"Defines the nonextensive deformed exponential used in the S-shaped curve.","marker":"[7]"},{"why":"Defines the deformed functions $\\ln_p$ and $\\exp_p$ that carry the model's analytical expressions.","marker":"[9]"},{"why":"Supports the ruin-aversion interpretation for the extreme-loss result.","marker":"[13]"},{"why":"Provides the contrast-ratio formula in decibels that the paper adapts to bound the fuzzy region.","marker":"[10]"}],"fun_headline_variants":["People pick sure bets over extreme odds, new experiment shows","Risk taking follows time-average wealth, not expected value","Extreme outcomes reveal a bias for safe growth over lottery","Prospect Theory challenged: time-average growth wins","In gambles, people favor time-average wealth, study finds"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The key assumption is that respondents' choices in the three problems came from thinking of the gambles as repeated many times and comparing long-run growth, not from the sheer size of the amounts, simple loss aversion, or a one-shot fear of ruin.","fun_headline_variants_meta":{"raw":{"variants":["People pick sure bets over extreme odds, new experiment shows","Risk taking follows time-average wealth, not expected value","Extreme outcomes reveal a bias for safe growth over lottery","Prospect Theory challenged: time-average growth wins","In gambles, people favor time-average wealth, study finds"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000268,"raw_usage":{"total_tokens":1566,"prompt_tokens":838,"completion_tokens":728,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":454,"completion_tokens_details":{"reasoning_tokens":649}},"tokens_in":454,"tokens_out":728,"duration_ms":7514,"temperature":1.0,"reasoning_tokens":649,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T15:06:19.182306+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Present the same three problems to two groups: one told the gamble will occur exactly once, the other told it will repeat many times. If the certainty-choice rates are equally high in both groups, the paper's time-average mechanism fails; if the repeated frame raises certainty choices above the one-shot frame, the central claim is supported.","supporting_citations":[{"cited_title":"Prospect theory: An analysis of decision under risk","cited_arxiv_id":null,"evidence_quote":"Provides the original Prospect Theory formulation and the experimental paradigm whose risk-seeking patterns this paper challenges."},{"cited_title":"Advances in prospect theory: Cu- mulative representation of uncertainty","cited_arxiv_id":null,"evidence_quote":"States that risk seeking holds only when outcomes are not extreme, the exact gap this paper targets."},{"cited_title":"Evaluating gambles using dynamics","cited_arxiv_id":null,"evidence_quote":"Supplies the time-average growth-rate analysis for repeated gambles that the model adopts."},{"cited_title":"Possible generalization of boltzmann-gibbs statis- tics","cited_arxiv_id":null,"evidence_quote":"Defines the nonextensive deformed exponential used in the S-shaped curve."},{"cited_title":"General- ized algebra within a nonextensive statistics","cited_arxiv_id":null,"evidence_quote":"Defines the deformed functions $\\ln_p$ and $\\exp_p$ that carry the model's analytical expressions."},{"cited_title":"Skin in the Game: Hidden Asymmetries in Daily Life","cited_arxiv_id":null,"evidence_quote":"Supports the ruin-aversion interpretation for the extreme-loss result."},{"cited_title":"Re- generative and reconfigurable all-optical logic gates for ultra-fast appli- cations","cited_arxiv_id":null,"evidence_quote":"Provides the contrast-ratio formula in decibels that the paper adapts to bound the fuzzy region."}],"review_version":1}