Deciphering the News Narrative: The VERA-J Framework in Action

Overview

Sometimes, media reports scientific research to the public in ways that, whether unknowingly or with the underlying intent to generate traffic, misstate a study's original intent. Whether via mainstream news, social media posts, or even university public relations offices, magnified findings or missing content can distort public perception and encourage actions or social initiatives that are unsupported by the actual research study. 

To mitigate potential distortion, we are "experimenting" with the VERA-J Agent framework 2.1, a tool we designed to systematically analyze and evaluate the reliability of media reports, opinion pieces, and news stories that are occasionally reported with chosen angles or emphases. The agent is a modification of our VERA research agent model, designed to asist assess research papers for the general public.

Below, we explain the article's why, detail the framework's mechanics, present its generated outputs, and demonstrate its application in a real-world setting using a Washington Post article. Also, for those unfamiliar with coding, we provide a prompt for major AI platforms that produces the VERA-J outcome.

Context

Recently, the Washington Post (WP) published a news report titled "Midlife fitness linked to longer, healthier lives," based on a study published in the Journal of the American College of Cardiology (JACC) titled "Midlife Cardiorespiratory Fitness and Healthy Aging: An Observational Cohort Study."

To assess the WP report's accuracy, as an example, we use the VERA-J agent framework, which also includes a subcomponent tailored for deep, politically sensitive, scientific, legal, financial, or policy-related pieces. This process aims to instill confidence, foster critical thinking, and reinforce the importance of scientific rigor for the general public audience.

As an operating standard in our work, we regularly conduct media scrutiny, with particular attention to research studies circulated across different media platforms. Our assumption is that they often influence public discussions, health choices, and policy debates to some degree, depending on the depth of knowledge of the population exposed to it.

When research studies are filtered through intermediaries such as popular news outlets, social media, or university press offices, the risk of unintentional misrepresentation increases. In the book Wrong Number: How to Extract Truth From a Blizzard of Quantitative Disinformation, Aaron Brown discusses methods for diagnosing statistical errors in journal articles and media accounts to minimize distortions in public perception caused by misrepresentations, missing context, or erroneous cause-and-effect relationships.

In our research, we notice, for instance, that if the quantitative aspect of a study is given greater weight in reporting while the qualitative aspect is inadequately presented, this can lead to false generalizations. These observations led us to develop the agent and reaffirmed our responsibility to ensure accuracy and clarity when introducing articles to our audiences, especially when supporting students and nonprofit organizations in their work. 

At the end, you will see that running the agent on the Washington Post piece about midlife fitness, the framework highlighted that even a fundamentally strong article can still fall short on representativeness, leaving the public with an incomplete map of the evidence.

VERA-J Framework Assessment Agent

Purpose:

The VERA-J Agent evaluates newspaper articles, magazine articles, online news reports, opinion pieces, and media analyses by examining source credibility, claim quality, evidence, framing, missing context, causal reasoning, and corroboration.

VERA-J Mechanics

V — Verify the Source: The agent first verifies the article's basic credibility.

It asks:

Who published the article?

Who wrote it?

Is the article news, analysis, opinion, editorial, sponsored content, or commentary?

Does the outlet have editorial standards and a corrections policy?

Is the author a beat reporter, columnist, expert, freelancer, or anonymous writer?

Is the article current, updated, corrected, or retracted?

Are original sources linked or named?

Are there obvious conflicts of interest, advocacy positions, or institutional biases?

Goal: 

Determine whether the article comes from a source that deserves initial trust, caution, or skepticism.

E — Explore the Claim

The agent identifies what the article is actually asking the reader to believe.

It separates: main claim, secondary claims, factual claims, interpretations, assumptions, speculation, emotional framing, and headline framing.

It asks:

What is the core claim?

Does the body of the article support the headline?

What facts are directly reported?

What is interpretation or inference?

What is implied but not proven?

Are terms clearly defined?

Are numbers or examples presented with context?

Goal: 

Convert the article from a narrative into a claim map.

R — Review the Evidence

The agent evaluates the quality of the evidence behind the claims.

It asks:

Does the article rely on primary documents?

Does it cite court filings, government data, reports, research studies, official statements, or direct interviews?

Are sources named or anonymous?

Are anonymous sources used responsibly?

Are multiple viewpoints included?

Are numbers contextualized?

Are opposing explanations considered?

Is the evidence sufficient for the article's conclusion?

Is the article cherry-picking examples?

Does it confuse anecdote with trend?

Goal: 

Determine whether the article's claims are supported by reliable evidence.

A — Assess the Interpretation

The agent judges whether the article's conclusion is warranted.

It asks:

Does the article overstate certainty?

Does it confuse sequence with causation?

Does it confuse correlation with causation?

Does it generalize from a single case to a broader trend?

Does it omit important context?

Does it fit with other credible reporting, public records, research, or official data?

What can reasonably be concluded?

What remains uncertain?

Goal: 

Produce a disciplined interpretation of what the article allows us to say and what it does not allow us to say.

VERA-J News Article Assessment Output

When you run (input) an article, the agent will produce the following:

1. Citation and Source Check

Identify: 

article title, author, publication, date,

article type:

news, analysis, opinion, editorial, commentary, sponsored content, or unclear, publication credibility, 

author: 

credibility, corrections, updates, or retractions, and whether the article links or cites sources.

2. Main Claim

State the article's central claim in plain language.

Also, identify whether the headline is: accurate, overstated, incomplete, misleading, or unsupported by the

article body.

3. Claim Map

Separate the article into:

Type of Statement Examples

Factual claims: Directly verifiable statements

Interpretive claims: The author's explanation of meaning

Causal claims: Claims that one thing caused another

Predictive claims: Claims about what may happen

Normative claims: Claims about what should happen

Unsupported implications: Ideas suggested but not proven

4. Evidence Used

Identify the evidence base:

documents, data, interviews, expert quotes, official statements, unnamed sources, research studies, court records, anecdotes, historical comparisons, or prior reporting.

5. Source Quality

Evaluate whether the article relies on

primary sources, named sources, qualified experts, independent verification, multiple perspectives, anonymous sources, partisan sources, advocacy organizations, or secondhand reporting.

6. Framing and Language

Assess whether the article uses:

neutral language, emotionally loaded language, selective framing, dramatic wording, vague attribution, misleading images, unsupported labels, or a narrative structure that pushes the reader toward a conclusion.

7. Correlation vs. Causation

State whether the article properly distinguishes:

timing, association, responsibility, causation, intent, effect, and consequence.

Flag any case where the article implies causation without sufficient evidence.

8. Missing Context

Identify what the reader would need to know to interpret the article fairly, such as:

historical background, baseline data, comparison groups, alternative explanations, legal or policy context, economic context, scientific context, counterevidence, or relevant limitations.

9. Strengths

Identify what the article does well, such as:

strong sourcing, primary documents, balanced perspectives, clear timeline, careful language, useful data, relevant expert input, or appropriate caveats.

10. Limitations and Red Flags

Identify weaknesses such as:

one-sided sourcing, anonymous sourcing without explanation, no primary documents, headline overreach, unsupported causal claims, missing data, vague claims, misleading comparisons, opinion presented as fact, outdated information, or lack of corroboration.

11. Fit With Other Evidence

Assess whether the article is:

corroborated by other credible reporting, supported by public records, consistent with official data, consistent with academic research, contradicted by other credible evidence, or too isolated to support strong conclusions.

12. Bottom-Line Judgment

Use one of the following ratings:

Strong article: Well-sourced, balanced, evidence-backed, and appropriately cautious.

Moderate article: Useful, but has some sourcing, framing, or context limitations.

Weak article: Thinly sourced, overstated, one-sided, or insufficiently contextualized.

Unreliable article: Major factual, sourcing, framing, or evidentiary problems.

13. Practical Interpretation

End with:

What this article allows us to say

State the conclusions that are reasonably supported.

What this article does not allow us to say

State the conclusions that would be overreach.

Real World Application

VERA-J Agent News Article Assessment Output

Article assessed:

Gretchen Reynolds. "Midlife fitness linked to longer, healthier lives, study finds." The Washington Post, May 17, 2026.

The article reports on the JACC study "Midlife Cardiorespiratory Fitness and Healthy Aging: An Observational Cohort Study," which examined whether midlife cardiorespiratory fitness was associated with later-life health span, disease burden, and lifespan among participants in the Cooper Center Longitudinal Study.

1. Citation and Source Check

Publication: The Washington Post, Well+Being / Fitness.

Author: Gretchen Reynolds.

Article type: Health journalism/science reporting, not an editorial or opinion article.

Date: May 17, 2026.

Underlying evidence: A peer-reviewed JACC observational cohort study using Cooper Center Longitudinal Study data linked to Medicare administrative claims.

The article is credible in several respects: it reports on a peer-reviewed study, identifies the main result, uses cautious "linked to" language, and acknowledges that the study is associational rather than causal. The author is an experienced exercise science journalist, and the article's topic falls within her reporting domain.

VERA-J source judgment: Strong source, but the article should have more explicitly explained the demographic and selection limits of the underlying study population.

2. Main Claim

The article's main claim is that higher cardiorespiratory fitness in midlife is associated with longer life and more years lived free of major chronic disease.

The headline: "Midlife fitness linked to longer, healthier lives, study finds," is generally accurate. It uses the careful phrase "linked to," which correctly signals association rather than proof of causation.

However, the article's public-facing framing could lead some readers to interpret the findings too broadly unless the demographic limitations are made clear. The finding is strongest for the specific cohort studied: people from the Cooper Center Longitudinal Study who underwent preventive exams, remained apparently healthy through age 65, and were later linked to Medicare fee-for-service claims.

3. Claim Map

Factual claim: The underlying study included 24,576 participants from the Cooper Center Longitudinal Study and linked them to Medicare claims. Supported.

Factual claim: Participants' midlife cardiorespiratory fitness was assessed before age 65 and categorized into age- and sex-adjusted fitness groups. Supported.

Factual claim: The study examined health span, disease burden, and lifespan. Supported.

Interpretive claim: Being more fit in midlife may be related to healthier aging. Reasonable.

Causal claim: The article mostly avoids claiming that fitness causes a longer health span. Appropriate.

Predictive claim: A brisk walk today may help future health span. Plausible, but stronger than the study alone can prove.

Unsupported implication risk: Readers may infer the result applies equally to all racial, socioeconomic, geographic, and health-status groups. That is not established.

4. Evidence Used

The article relies mainly on:

The JACC observational cohort study.

Cooper Center Longitudinal Study preventive-exam data.

Medicare administrative claims.

Cardiorespiratory fitness measurement from treadmill testing.

Expert interpretation from study authors and at least one outside expert, according to the article summary.

The underlying JACC study's stated aim was to examine associations between midlife cardiorespiratory fitness and later-life health span, disease burden, and lifespan among adults who remained apparently healthy through age 65.

Evidence judgment:

Good evidence for an observational health-news article, but not definitive causal evidence.

5. Source Quality

The source quality is moderately strong to strong.

The article has several strengths:

It is based on a peer-reviewed article in a cardiovascular journal.

The study sample is large.

Fitness was measured using a treadmill-based clinical assessment rather than relying solely on self-reported exercise.

Outcomes were linked to Medicare claims, which are more objective than self-reported disease history alone.

The article uses cautious language about association.

But the evidence source has limitations:

The study is observational.

The cohort is not a random sample of the U.S. population.

Participants came from a preventive medicine clinic population in Dallas, Texas.

The study included only people who remained apparently healthy through age 65.

Only 25% of participants were women.

Based on the study review, the cohort was predominantly White, which should have been stated more explicitly in the article.

The sample likely overrepresented health-conscious people who had access to preventive care and were able or motivated to attend Cooper Clinic exams.

6. Demographic Composition, Representativeness, and Generalizability

The article should have made it clearer that the underlying population was not broadly representative of the U.S. population.

The JACC study included 24,576 participants, but only 25% were women. The participants came from the Cooper Center Longitudinal Study, based on preventive medicine examinations at the Cooper Clinic in Dallas, Texas, from 1971 to 2017, and were later linked to Medicare fee-for-service claims from 1999 to 2019.

That means the cohort likely differs from the general population in several ways:

Representativeness factor and why it matters

Race/ethnicity: If the cohort was mostly White, the findings may not generalize to Black, Hispanic, Asian, Native American, or other populations with different baseline risks, access to care, structural health factors, and chronic disease patterns.

Sex: Only 25% of participants were women, limiting certainty about whether effect sizes apply equally to women.

Socioeconomic status: Preventive clinic populations often overrepresent individuals with greater resources, better insurance, higher health literacy, and greater access to care.

Geography: The clinic population came from Dallas, Texas; it is not necessarily representative of all U.S. regions or rural populations.

Health status: The study focused on adults who remained apparently healthy through age 65, excluding those with certain earlier diseases or risk factors.

Selection bias: People who seek preventive medical exams and undergo fitness testing may already differ in motivation, lifestyle, income, baseline health, and healthcare engagement.

VERA-J generalizability judgment:

The article's central finding is credible for the study population, but its generalizability to the entire U.S. adult population should be approached with caution. The article should have explicitly stated that the cohort was demographically and behaviorally selected, particularly with respect to race/ethnicity, sex balance, socioeconomic status, geography, and access to preventive care.

A stronger article would have said:

These findings come from a selected, largely health-conscious Cooper Clinic cohort, not a nationally representative sample. The results may not apply equally across racial, socioeconomic, sex, geographic, or medically underserved populations.

7. Framing and Language

The article's framing is mostly responsible.

Positive features:

The headline says "linked to" rather than "causes."

The article recognizes the study as an association study.

It avoids presenting the study as a randomized trial.

It communicates the difference between living longer and living healthier.

Potentially overbroad framing:

The practical takeaway may sound universal: exercise now, live longer and healthier later.

The article should have made this point more clearly: "in this selected cohort."

It should have made the limits of demographic composition and representativeness more visible rather than leaving them as implicit methodological details.

VERA-J framing judgment:

Mostly accurate, but slightly too universal in tone.

8. Correlation vs. Causation

The article does well here.

The Washington Post summary explicitly states that the study was associative and cannot prove that fitness was the primary driver of longer or healthier life. That is important because higher midlife fitness may be partly a marker for other factors:

genetics, lifelong health behaviors, income, education, access to healthcare, diet, occupational conditions, neighborhood environment,

lower baseline disease risk, or better healthcare engagement.

VERA-J causality judgment:

The article properly identifies the finding as an association. It should, however, have more explicitly linked the non-causal limitation to the cohort's demographic and socioeconomic selection.

9. Missing Context

The article should have added more context in at least six areas.

First, it should have explicitly stated that the study population was not nationally representative.

Second, it should have reported key demographic composition more visibly, especially the low proportion of women and, based on your observation, the mostly White racial composition.

Third, it should have explained the Cooper Clinic selection issue: participants in the preventive medicine clinic are not representative of the general population.

Fourth, it should have clarified that "fitness" is not identical to "exercise." Cardiorespiratory fitness is influenced by physical activity, genetics, health status, body composition, prior disease, and other factors.

Fifth, it should have explained whether fitness was measured once or repeatedly. A single midlife measurement cannot fully capture lifetime fitness trajectories. The ScienceDirect abstract notes that future work should examine changes in fitness over the life course.

Sixth, it should have explained which confounders were adjusted for and which may remain unmeasured.

10. Strengths

The article has several strengths.

It reports on a large cohort study rather than a small anecdotal sample. It uses a clinically meaningful outcome: health span, disease burden, and lifespan. It correctly frames the finding as a link, not as proof of causation. It also reports a study that used objectively assessed cardiorespiratory fitness rather than relying solely on self-reported exercise.

The article also clearly communicates its public-health relevance. For a general audience, the distinction between lifespan and health span is useful and appropriate.

11. Limitations and Red Flags

There are no major red flags suggesting the article is unreliable. The concern is not that the article is wrong; it could be more precise.

Key limitations:

The article reports an observational study.

The cohort was selected, not nationally representative.

Only 25% of participants were women.

Based on the reading, the cohort was mostly White.

The study involved participants from the preventive clinic in Dallas, Texas.

The participants were healthy enough to remain apparently disease-free through age 65.

Fitness may function partly as a proxy for other advantages.

The article could leave readers with the impression that the results apply broadly and evenly across the population.

VERA-J red flag level:

Low for factual reliability; moderate for generalizability and representativeness.

12. Fit With Other Evidence

The article fits well with broader evidence that cardiorespiratory fitness is associated with lower disease burden and mortality risk. The JACC study itself states that it examined midlife cardiorespiratory fitness in relation to later-life health span, disease burden, and lifespan, and found that higher fitness was generally associated with later onset of major disease and fewer conditions within disease groups.

The claim is biologically plausible and consistent with the broader exercise-science literature. However, the exact effect size — such as how many additional healthy years may be associated with higher fitness — should be treated as cohort-specific unless replicated in more diverse populations.

VERA-J corroboration judgment:

The broad association is well supported; the precise generalization to all groups is less certain.

13. Bottom-Line Judgment

Rating: Strong article, with an important caveat about generalizability.

The article is credible, responsibly written, and mostly careful with causal language. It accurately reports that higher midlife cardiorespiratory fitness was associated with longer health span and lifespan in a large observational cohort.

However, the article should have made the limits of the study population's representativeness more explicit. The underlying cohort was not a demographically balanced national sample. It was drawn from Cooper Clinic preventive-exam participants, included only 25% women, and, based on your reading of the study, was mostly White. Those factors matter because health, fitness, disease onset, and access to preventive care vary by race, sex, socioeconomic status, geography, and healthcare access.

A more precise bottom line would be:

This is a credible article about a credible observational study. Still, its public-facing interpretation should be limited to the selected cohort studied and should not be presented as equally generalizable to all demographic groups.

What this article allows us to say

This article shows that, in a large but selected observational cohort, adults with higher midlife cardiorespiratory fitness were associated with a later onset of major chronic disease, a longer period of life without major disease, a lower later-life disease burden, and a modestly longer lifespan.

It also allows us to say that maintaining or improving cardiorespiratory fitness in midlife is consistent with broader evidence on healthy aging. However, the exact magnitude of benefit may differ across populations.

What this article does not allow us to say

This article does not allow us to say that fitness alone caused the longer health span or longer lifespan observed in the study.

It also does not allow us to say that the findings apply equally to all racial, ethnic, sex, socioeconomic, geographic, or medically underserved groups. The underlying study population was selected from preventive clinic participants, was mostly White, and included only 25% women.

Finally, it does not specify the exact exercise program, intensity, frequency, or duration required to produce the observed association across different populations.

Prompt to use the VERA-J agent

You are an expert media analyst and fact-checker who applies the VERA-J Framework to evaluate journalism, news articles, opinion pieces, and social media claims. VERA-J has four phases. Work through every phase systematically.

VERA-J is designed for content that is NOT a scientific research paper. If the content IS primarily a research paper, note this and recommend applying the VERA Research Framework instead.

PHASE 1 — VERIFY: Outlet & Author Credibility

Step 1a — Outlet classification:

 Classify the publication or platform into one of these types:

   LEGACY PRESS — Established newspaper or magazine with editorial standards,

                        named masthead, corrections policy (e.g., NYT, WSJ, BBC, Reuters).

   DIGITAL NATIVE  — Online-only outlet founded with journalism as its mission

                        (e.g., The Atlantic online, Vox, ProPublica, The Intercept).

   PARTISAN / ADVOCACY — Outlet with a disclosed or well-documented political or

                        ideological orientation that shapes story selection and framing.

   TRADE / SPECIALTY — Outlet serving a professional or industry audience; may be

                        credible within its domain but lacks general editorial oversight.

   TABLOID / ENTERTAINMENT — Prioritizes engagement over rigor; sensational framing is common.

   CONTENT FARM — High-volume low-verification content; often SEO-driven.

   FRINGE / CONSPIRATORIAL — Documented pattern of publishing false or misleading content.

   SOCIAL MEDIA POST — Individual user post; no editorial oversight by default.

   UNKNOWN / CANNOT ASSESS — Outlet not in training knowledge; flag for manual check.

  If the outlet is UNKNOWN or CANNOT ASSESS, instruct the reader to verify at:

   • mediabiasfactcheck.com

   • allsides.com

   • Ad Fontes Media (adfontesmedia.com)

Step 1b — Editorial independence:

 Is the outlet independently owned, or part of a larger media conglomerate?

 Is there a known ownership interest, advertiser relationship, or institutional

 affiliation that could shape coverage of this topic?

 Flag any known conflicts between the outlet's financial interests and the story's subject.

Step 1c — Author credentials:

 Is the author identified by name? If anonymous: flag this. Anonymous authorship

 removes a key accountability mechanism.

 Is the named author a trained journalist, a subject-matter expert, a public figure,

 or an unknown individual?

 Does the author have a track record covering this beat, or is this outside their domain?

Step 1d — Original reporting vs. aggregation:

 Is this original reporting (the journalist interviewed sources, obtained documents,

 or conducted direct observation)? Or is it aggregating / republishing content from

 another outlet? If aggregated, name the original source where possible.

Step 1e — Date and timeliness:

 When was this published? Is the content being shared now significantly older

 than the publication date (recycled or "zombie" content)?

 Is the topic one where a 6–12-month-old article may be materially outdated? 

PHASE 2 — EXAMINE: Claim Identification

Step 2a — Central claim:

 State the primary claim being made in one clear sentence. This is what the

 piece is asserting to be true. Do not use the headline — extract the actual

 claim from the body of the content.

Step 2b — Claim type:

 Classify the claim into one or more of these types:

   FACTUAL— Asserts something happened or is verifiably true/false.

   CAUSAL — Asserts that X caused, produced, or led to Y.

   PREDICTIVE — Asserts that something will happen in the future.

   VALUE/OPINION — Asserts that something is good, bad, fair, or should be done.

   INTERPRETIVE — Asserts a meaning or significance that depends on framing.

 This matters because factual claims can be checked; value claims cannot.

 Flag whenever a value or opinion claim is presented as if it were factual.

Step 2c — Falsifiability:

 What evidence would prove this claim wrong? If no evidence could falsify the claim,

 flag it as unfalsifiable — a structural feature of propaganda and motivated reasoning.

Step 2d — Source chain depth:

 Trace how far the claim is from its original source:

   PRIMARY — Journalist spoke directly to the source / obtained the document firsthand.

   SECONDARY — Journalist is reporting on another outlet's reporting.

   TERTIARY+ — The claim has passed through multiple intermediaries.

 Each step away from the primary source increases the risk of distortion.

 If the article cites another article, which cites another article — flag the chain.

Step 2e — Headline vs. body alignment:

 Does the headline accurately represent what the body actually says?

 Headlines are read far more often than articles. A misleading headline is a

 significant accuracy failure, even if the body is careful.

 Classify the alignment:

   ACCURATE — Headline is a fair representation of the body.

   OVERSTATED — Headline makes a stronger claim than the evidence in the body supports.

   MISLEADING — Headline implies something the body contradicts or does not support.

   CLICKBAIT — Headline withholds key context to manufacture curiosity or alarm.

PHASE 3 — REVIEW: Evidence & Sourcing Quality

 3A — Evidence Hierarchy Assessment

Identify what type(s) of evidence the piece relies on, using this hierarchy

from strongest to weakest:

 TIER 1 — Primary documents, official data, or on-record expert statements

           with verifiable credentials. (Government records, peer-reviewed data,

           named expert quoted in their field of expertise.)

 TIER 2 — Named expert sources outside their primary field, or named

           non-expert sources with direct relevant experience (witnesses,

           affected individuals, practitioners).

 TIER 3 — Multiple anonymous sources corroborating the same claim independently.

 TIER 4 — Single anonymous source.

 TIER 5 — Unattributed assertions: "experts say," "studies show," "sources claim,"

           "many people believe" — with no named source or citation.

  State the highest tier of evidence present AND the lowest tier relied upon

 for the central claim. Flag if the central claim rests primarily on Tier 4 or 5.

3B — Expert vetting:

 Are the cited experts genuinely experts in the domain relevant to the claim?

 A cardiologist commenting on vaccine immunology, or an economist commenting

 on clinical trials, is an expert outside their domain — flag this.

 Does the expert have a disclosed financial or ideological interest in the outcome?

3C — Counterevidence and balance:

 Does the piece acknowledge opposing evidence, dissenting experts, or

 alternative interpretations?

 If only one side is presented, is this because no credible opposition exists

 (appropriate) or because the opposition was ignored (problematic)?

 Flag false balance: giving equal weight to a fringe position and a scientific consensus

 is as misleading as ignoring dissent entirely.

3D — Red flags checklist:

 Flag any of the following that appear in the content:

 ⚠ Vague citations  — "studies show," "research suggests," "experts agree" without

                       naming the study, researchers, or experts.

 ⚠ Appeal to authority — citing credentials without relevance to the specific claim.

 ⚠ Anecdote as evidence — treating one person's experience as representative data.

 ⚠ Emotional amplification — language designed to produce fear, anger, or outrage

                             rather than to inform.

 ⚠ Strawman opposition — misrepresenting the opposing view before dismissing it.

 ⚠ Omitted denominator — reporting a number without the base rate that gives it meaning

                          (e.g., "100 cases reported" without stating "out of 10 million").

 ⚠ Post hoc reasoning — inferring causation from sequence ("X happened, then Y happened,

                         therefore X caused Y").

 ⚠ Overgeneralization — applying findings from a narrow group to all people.

 3E — If a study IS cited:

 Is the study named specifically enough to verify (journal, authors, year)?

 Or is it cited only vaguely ("a study from Harvard," "recent research")?

 If specific, note whether the claim accurately represents what the study found.

 If vague: flag that the claim cannot be independently verified.

PHASE 4 — ASSESS: Framing, Context & Accuracy

4A — Language audit:

 Audit the language for framing choices that shape interpretation beyond the facts:

 • Causal vs. correlational language: Does the piece assert causation for

   a relationship that is only correlational? ("X causes Y" vs. "X is linked to Y")

 • Loaded language: Words chosen for emotional impact rather than precision.

 • Passive construction used to obscure agency: "Mistakes were made" — by whom?

 • Weasel words: "Some say," "critics argue," "many believe" — who exactly?

 • Certainty inflation: Presenting an emerging or contested finding as a settled fact.

4B — Missing context:

 What information, if added, would materially change how a reader interprets the claim?

 Common omissions:

   • Base rates (how common is this normally?)

   • Comparison groups (compared to what?)

   • Time horizon (over what period?)

   • Funding or interest of the person making the claim

   • What the other side of the story actually argues

 Flag each significant omission explicitly.

4C — Scope and generalizability:

 Does the piece apply findings or events from a narrow context to a broad population?

 Example: a policy outcome in one city presented as a lesson for all cities;

 A health outcome in one demographic is presented as universal.

 Flag any scope creep as a significant framing problem.

4D — Conflict of interest:

 Does the author, outlet, or any cited source have a financial, political, or

 personal stake in the claim being accepted as true?

 Conflicts of interest do not invalidate content — but they must be disclosed and weighed. 

4E — Verification roadmap:

 Provide 2–4 specific, actionable steps the reader can take to independently

 verify the central claim. Be specific: name the databases, agencies, or

 organizations to consult for this particular topic.

4F — Overall credibility rating:

 Assign ONE rating with a one-paragraph justification:

 RELIABLE — Named sources; primary documents linked or cited; claims match

                    evidence presented; headline accurately represents body; outlet

                    has a visible corrections policy; no significant red flags.

 MOSTLY RELIABLE  — Minor sourcing gaps; mostly named sources; headline slightly

                    overstates but body is accurate; outlet has editorial standards;

                    No major red flags; suitable for reference with light verification.

 USE WITH CAUTION — Anonymous sourcing for key claims; missing counterevidence;

                    headline overstates or misleads; outlet with documented partisan

                    lean; emotional amplification present; central claim not

                    independently verifiable from the piece alone.

 UNRELIABLE — Fabricated, unverifiable, or demonstrably false claims; headline

                    directly contradicts body; known misinformation outlet; no

                    corrections policy; vague citations throughout; multiple Tier-5

                    evidence failures; significant red flags from the checklist.

OUTPUT FORMAT — use this exact structure every time

Content Summary

[1–2 sentence plain-language description of what is being claimed and by whom]

Phase 1 — VERIFY

[Outlet classification; editorial independence; author credentials; original vs.

aggregated; date and timeliness. Flag any ⚠ prominently.] 

Phase 2 — EXAMINE

[Central claim; claim type; falsifiability; source chain depth; headline vs.

body alignment with classification]

Phase 3 — REVIEW

[Evidence hierarchy (tiers present; tier of central claim);

expert vetting; counterevidence and balance;

red flags checklist (flag each triggered item with ⚠);

study citation quality if applicable]

Phase 4 — ASSESS

[Language audit; missing context (list each omission);

scope and generalizability; conflict of interest;

verification roadmap (2–4 specific steps);

overall credibility rating with justification paragraph]

Bottom Line

[2–3 sentences: what the content actually claims, how reliable the sourcing is,

what the reader should do before acting on or sharing this information]

STANDING RULES

• Distinguish factual claims from value claims. Do not fact-check opinions.

• Distinguish the outlet's credibility from the truth of the specific claim —

 a credible outlet can publish an inaccurate story; a low-credibility outlet

 can accidentally publish something true.

• Never fabricate source assessments or outlet histories not present in your

 training knowledge. If you do not have reliable information about an outlet

 or author, say so explicitly and provide manual verification steps.

• Flag every gap explicitly rather than skipping sections with missing information.

• Never refuse to analyze because information is incomplete — analyze what is

 present and flag every gap.

• Do not conflate partisan lean with unreliability. A partisan outlet can still

 publish accurate, well-sourced content. Assess sourcing and evidence, not politics.

• Assess the content's claims, not whether you agree with the conclusion.

 Your role is to evaluate the evidence and sourcing, not to render a verdict

 on the underlying policy or factual dispute.

YOUR CONTENT TO ANALYZE

[PASTE OR TYPE YOUR NEWS ARTICLE, SOCIAL MEDIA POST, HEADLINE, OR CLAIM BELOW THIS LINE]

Finally

In an information environment where multiple picks, valleys, and holes are present, and detailed scientific methodologies are routinely reduced to simplified public narratives, structured evaluation is a must. The VERA-J Agent framework provides a mechanism for this process. Running the agent on the Washington Post piece about midlife fitness, the framework highlighted that even a fundamentally strong article can still fall short on representativeness, leaving the public with an incomplete map of the evidence.

By systematically verifying sources, exploring claims, reviewing evidence, and assessing interpretations, VERA-J goes into the narrative framing to isolate the actual data. Ultimately, the goal is not to negate health and science journalism but to advance how we interact with it. By suggesting the use of the agent, we aspire to empower readers, students, and policymakers to engage with media critically, helping to ensure their decisions are firmly grounded in research rather than the appeal of headlines.

Remember, this is an ongoing experiment.

Sources

Sumner, P., Vivian-Griffiths, S., Boivin, J., Williams, A., Venetis, C. A., Davies, A., Ogden, J., Whelan, L., Hughes, B., Dalton, B., Boy, F., & Chambers, C. D. (2014). The association between exaggeration in health-related science news and academic press releases: retrospective observational study. BMJ (Clinical research ed.), 349, g7015. https://doi.org/10.1136/bmj.g7015

Universiteit Leiden. (2018, January 2). Exaggeration in medical news starts with the press release. https://www.universiteitleiden.nl/en/news/2018/01/exaggeration-in-medical-news-starts-with-the-press-release

Notes

(A)  A group of friends from a private network of current and former team members from equity firms, entrepreneurs, Disney Research, and universities such as NYU, Cornell, MIT, Eastern University, and UPR gather to share articles and studies based on their experiences, insights, inferences, and deductions, often using AI platforms to support research and communication.