Regulatory risk assessments commonly use conservative assumptions that overstate risk
6 events · 2 assessments · 3 decisions
Reassessed (status unchanged)
Triggered by the contradicts-subclaim on offsetting non-conservative assumptions receiving its first assessment (SUPPORTED, 0.68, modal reading; strong reading contested). Materiality check: this standing is precisely what the prior assessment already assumed for that premise, so the change is absorbable without a status change. The firmed-up against-line only defeats the strong reading (large systematic overstatement of true risk), which the verdict already declined to endorse; it does not push the central health-protective case toward net understatement. Kept status SUPPORTED (confidence 0.72, credence 0.75), refreshed the assessment prose to reference the now-assessed premises inline, and re-confirmed both argument evaluations (both holds_with_caveats) against current premise standings. Importance left at 0.4 (notable within regulatory-policy methodology; moderate contestation). No dependent notification: no material change to this claim's status or credence.
Reassessed: still Supported
verdict confidence 0.72 · credence 0.75
No action needed
Trigger: subclaim_change on c6f0b7ce ('Compounding multiple conservative assumptions causes regulatory risk estimates to substantially exceed actual risk'), which recorded its first assessment as CONTESTED (confidence 0.8): the compounding mechanism is well supported, but the strong claim of substantial exceedance of TRUE (unknown) risk is blocked by the unmeasurable true-risk gap and offsetting non-conservative assumptions. Materiality judgment: not material to this claim's status. The parent's current assessment (supported, conf 0.72, credence 0.75) was already written on the modest reading and explicitly anticipated this subclaim being contested ('which is why the crux subclaim is marked contested'). The parent turns on the compounding MECHANISM plus the deliberate-conservatism descriptive premise producing upper-bound rather than central estimates, not on the strong quantified claim about true risk that the subclaim disputes. The subclaim moving from unassessed to formally CONTESTED confirms, rather than disturbs, the existing verdict and its argument evaluation ('holds_with_caveats', securing the modest but not the strong conclusion). Coherence check (§21): a 'supports' subclaim being contested in its strong form is consistent with the parent being supported in its modest form. No status, confidence, credence, canonical-form, structure, or argument-evaluation change warranted. Importance left at 0.4 (notable empirical claim about regulatory practice; moderate contestation). No dependent notification, since nothing changed here.
Structured and assessed
First pass. Decomposed into three novel subclaims (Matcher confirmed none existed): a descriptive premise that agencies deliberately adopt conservative health-protective defaults (low importance 0.18, settled, left as embedded stub), a contested causal crux that compounding conservative assumptions substantially overstates true risk (0.4, contestation 0.7), and a counter-consideration that offsetting non-conservative assumptions can reverse net bias (0.38, contestation 0.65). Grouped the first two under a 'for' argument and the third under an 'against' argument; wrote both forms and evaluated both (holds_with_caveats). Grounded via web search in the EPA/NRC risk-assessment literature (plausible conservatism, compounding conservatism) and the Finkel-style critique. Set importance 0.4 (notable-to-major, domain-confined but genuinely contested and structurally load-bearing as the against-line for claim 638a73b7). Assessed SUPPORTED, confidence 0.72, credence 0.75: the modest reading is favored (conservative assumptions common and upward-biasing), while net magnitude/systematicity of overstatement of true risk is unresolved; not verified for that reason, not contested because the core proposition is accepted across the debate. Flagged the domain distinction from the understatement genre (nuclear PRA, VaR). Notified the one dependent steward.
Assessed Supported
verdict confidence 0.72 · credence 0.75
In the domain this claim describes, chiefly chemical, environmental, and public-health risk assessment as practiced by agencies such as the U.S. Environmental Protection Agency, it is well established that assessments deliberately build in conservative, health-protective assumptions: upper-bound (rather than central) cancer potency estimates, reasonable-maximum-exposure scenarios, uncertainty and safety factors in reference doses, and non-threshold dose-response models. Because these inputs are chosen to err on the side of caution, and because several are typically combined in a single estimate, the resulting figures generally lie above best-estimate (central) risk and, in that sense, overstate risk. Agencies themselves describe their outputs as intentionally conservative, and past EPA reevaluations of specific estimates have generally revised the numbers downward, consistent with prior overstatement relative to central estimates. What remains genuinely disputed is the strong version of the claim: how large the net overstatement of true risk is, and whether it is systematic. Critics of the "too conservative" narrative argue that assessments also contain non-conservative choices and omissions (unmodeled exposure pathways, susceptible subpopulations, untested endpoints, ignored mixtures) that can partly or wholly offset the conservative bias, so that the net direction is not guaranteed. No systematic calibration of regulatory estimates against realized outcomes exists to settle the magnitude. The evidence therefore favors the claim on its modest reading, that conservative assumptions are common and push estimates upward, while leaving the size and universality of the net effect on true risk unresolved. The claim should not be conflated with the separate observation that some probabilistic engineering and financial risk models (nuclear PRAs, bank value-at-risk) have historically understated risk; those concern a different genre of analysis and a different failure mode.
Claim entered the graph