How the Media Gets It Wrong

The Bias Detection Toolkit

by Iran Prism 9 min read

The Tool

In February 2019, Thomas Erdbrink — the New York Times’s bureau chief in Tehran — published a feature under the headline “From Theocracy to ‘Normality.’”1 He described coffee shops, Western clothing, women in public spaces. Readers in New York and Washington saw a country evolving toward something familiar. The article passed every test a casual reader would apply: credible outlet, experienced correspondent, firsthand observation, measured tone. Months later, Erdbrink’s credentials were revoked. The Times waited four months to report his expulsion. And under the “normality” he described, the regime was building the apparatus that, in January 2026, would kill thousands in forty-eight hours.

The article was a filter, and the reader had no way to detect it. This toolkit contains the detection method.

Four audit sections, each taking a different angle on the same question: is this analysis testing reality, or protecting a framework? Think of it as a home inspector’s checklist. The seller wants you to see the kitchen. The inspector checks the foundation.

Audit 1: Source and Demographics

Before reading a single argument, check who wrote it and who they cite.

The Author Check: What is the author’s institutional affiliation? Is it a department with a known extreme Democrat-to-Republican faculty ratio — seventeen to one in History, more than twenty to one in Sociology? This does not invalidate the work. It tells you the professional ecosystem in which the work was produced and the incentive structure the author navigated to publish it.

The Citation Loop Check: Open the bibliography. Does it cite diverse viewpoints, or does every reference come from the same theoretical camp? If a paper on Iranian politics cites only Constructivists, it is building on a foundation that shares its assumptions. The citations look like evidence. They are echo.

The Funding Check: Is the funding source disclosed? Does the output align suspiciously well with the funder’s strategic interests? A report funded by an organization that advocated for the JCPOA, concluding that the JCPOA was successful, warrants exactly the same skepticism as a report funded by the defense industry concluding that military spending is insufficient.

Worked Example — A Brookings JCPOA Paper: In April 2015, a Brookings panel titled “Deal or No Deal?” featured analyst Suzanne Maloney stating it was “reasonable to suspect” that figures within Iran’s executive might “embrace more lenient terms… if they were in fact empowered to do so.”2 The “if empowered” caveat described a structural impossibility in a system dominated by the Supreme Leader — but in media translation, the nuance evaporated. What survived was the prediction of moderation. The panel was held at an institution where key analysts — Robert Einhorn, Richard Nephew — had moved from the State Department that negotiated the deal to the think tank that evaluated it. The revolving door created a structural disincentive to question the premises of engagement, because doing so would indict their own professional legacies.

Source audit result: Brookings-State Department revolving door, citation loop with the New York Times, funding aligned with engagement framework.

Audit 2: Lexical and Framing

Language reveals what argument cannot hide. Three tests.

The Voice Test: Are Western or American actions described in active voice (“The U.S. bombed the facility”) while adversary actions use passive voice (“Violence erupted,” “Rockets were fired”)? Active voice assigns agency and responsibility. Passive voice removes the agent, presenting violence as a natural phenomenon rather than a deliberate choice.

The test works in reverse too. Does the analysis use active voice for Iranian regime actions (“The IRGC killed protesters”) or passive voice (“Protesters were killed in clashes”)? “Clashes” implies symmetry — two sides fighting. The reality in January 2026: armed security forces with DShK machine guns firing into unarmed crowds.

Case study — Two Aerial Shootdowns: Coverage of the Soviet shootdown of Korean Air Lines Flight 007 in 1983 used moral language: “murder,” “cold-blooded.” Coverage of the American shootdown of Iran Air Flight 655 in 1988 used technical language: “tragedy,” “mistake,” “misidentification.”3 Both killed all passengers. The grammatical framing determined which was understood as crime and which as accident.

The Label Test: Does the author use “government” for some states and “regime” for others? “Government” implies legitimacy. “Regime” implies the opposite. If an author writes “the Syrian Government” but “the Israeli regime” — or vice versa — they are encoding a normative judgment as description. Watch for it in both directions. The inconsistency is the tell.

The Stop-Thought Test: Watch for adjectives that function as arguments: “imperialist,” “neocolonial,” “unprovoked.” When these words replace evidence — when “the imperialist intervention” substitutes for “the intervention, which had these specific characteristics of imperialism” — the adjective is doing the work that analysis should do.

Worked Example — A BBC Report on January 2026: The BBC initially described the January 2026 events as “unrest” and “protests” sparked by “economic conditions.” The framing removed political agency from protesters who chanted “Death to the Dictator” and “We don’t want the Islamic Republic.” It implied economic causes rather than political rejection of the system. And it used passive construction (“sparked by”) that obscured who was killing whom. The diaspora media — Iran International, IranWire — used “uprising,” “revolution,” and “massacre” within days. The terminological gap lasted weeks and delegitimized the political nature of the movement during its most critical phase.

Audit 3: Methodology

Three checks for how evidence is handled.

The Streetlight Check: Does the analysis criticize an open society using data that is simply unavailable or suppressed for the closed society being compared? A study documenting American drone strike casualties in forensic detail while citing “official” Iranian casualty figures from the same regime that arrested Emadeddin Baghi for trying to count the dead is not balanced analysis. It is a product of the information asymmetry between open and closed societies — and the analyst’s failure to flag it.

If you have ever noticed that your mental image of “which country commits the worst abuses” correlates perfectly with “which country’s abuses are most documented” — that is the Streetlight Effect operating in your own thinking. The countries that look worst are often the countries most willing to investigate themselves.

The Counterfactual Check: Does the analysis calculate the cost of action without estimating the cost of inaction? Critiques of sanctions catalog civilian suffering — rightly. But do they estimate what happens if sanctions are lifted and the IRGC captures the revenue? Critiques of military strikes catalog destruction — rightly. But do they estimate the trajectory of a regime accelerating nuclear enrichment, expanding proxy wars, and massacring its own citizens? The cost of inaction is never zero. An analysis that treats it as zero is advocacy dressed as scholarship.

The Casualty Verification Check: Does the analysis accept casualty figures from authoritarian state bureaus as “official data” while demanding forensic standards for independent estimates? During January 2026, major outlets presented the regime’s death toll of approximately three thousand alongside independent estimates exceeding thirty thousand — without weighting the proven mendacity of the former.4 The state deliberately destroys evidence, then the media waits for “official confirmation” that the state actively prevents.

The Verification Asymmetry: Victims’ accounts arrive wrapped in doubt: “We cannot independently verify.” “These claims are unconfirmed.” The regime’s line goes out clean — quoted confidently, treated as baseline. When one side gets caveats and the other gets a microphone, “balance” becomes geometry that favors the state.

Worked Example — A Sociology Paper on the 2026 Uprising: A paper attributing the January 2026 uprising to “economic grievance” and “the collapse of the social contract” — using passive voice for regime killings (“lives were lost in the ensuing violence”) — fails all three methodology checks. It applies the Streetlight Effect by using economic data while ignoring GAMAAN political aspiration data. It omits the counterfactual of what happens to ninety-three million people if they do NOT rise up against an entrenching regime. And it treats the regime’s casualty figures as one legitimate data point in a “contested” range rather than as the output of a state with every incentive and demonstrated history of falsification.

Audit 4: Logic and Argumentation

The final layer. Three tests for whether the analysis is scholarship or theology.

The Steelman Test: Does the author explain the opposing view in a way its proponents would recognize? If a paper argues that engagement with Iran failed because of American bad faith, does it first articulate the strongest case for why engagement failed because of the regime’s structural design? If a scholar cannot explain why Iranians might support Reza Pahlavi in a way that a monarchist would accept as fair — not agree with, but accept as a fair description of their reasoning — they have not understood the phenomenon. They have dismissed it.

The Moral Equivalence Check: Does the analysis conflate intent with outcome? Accidental collateral damage with subsequent investigation, admission, and compensation is not morally equivalent to deliberate targeting of civilian infrastructure as a strategy of war. A drone strike that kills civilians by error and an IRGC DShK machine gun pointed at an unarmed crowd are not the same category of event. Analysis that treats them as equivalent is not being balanced. It is being evasive.

The Falsifiability Check: Is the argument structured so that contradictory evidence could disprove it? If every piece of evidence — reform, repression, engagement, sanctions, protest, silence — confirms the same theory, the theory is unfalsifiable.

The Null Hypothesis Test: Before accepting any Iran analysis, ask one question: “What specific evidence would convince the author they are wrong?” If moderation proves reform is possible, and repression proves reformers were undermined by hardliners, and both confirm the engagement thesis — the framework has survived every possible outcome. Theories that cannot fail are not theories. They are articles of faith.

The Honest Ledger

This toolkit is not a weapon for dismissing scholarship you disagree with. It is a diagnostic for checking whether analysis has been captured by the structural forces documented in Why Your Iran Expert Might Be Wrong.

Not every paper from a History department with a seventeen-to-one ratio is biased. Not every passive-voice construction is evasion. Not every citation loop is a cartel. The toolkit identifies patterns — and patterns require judgment. A single red flag is a caution. Multiple red flags across multiple audit sections is a diagnosis.

The cost of biased analysis is not abstract. It is measured in lives — the fifteen hundred killed in November 2019 while experts still called Rouhani a “moderate,” the thousands killed in January 2026 while analysts debated whether to say “uprising.”5 Every distortion that softens the regime’s image, every euphemism that buries its violence, every citation loop that insulates the canon from the country — these are not academic failures. They are failures with human consequences. The people whose lives depend on Western clarity about Iran deserve analysis that has passed an inspection, not analysis that has merely passed through the gates.

The Four-Audit Checklist (at a glance):

  1. Source & Demographics — Who wrote it? Who do they cite? Who funded it? Revolving door?
  2. Lexical & Framing — Active/passive voice symmetric? Labels consistent? Adjectives doing the work of evidence?
  3. Methodology — Streetlight check? Counterfactual present? Casualty verification equal?
  4. Logic & Argumentation — Steelman test? Moral equivalence check? Falsifiability?

Run all four. If the analysis survives, trust it more. If it fails three or four — read something else.

The goal is not neutrality, which is often a mask for the status quo. It is objectivity: the disciplined adherence to evidence regardless of where it leads, and the courage to test assumptions against the country rather than the canon. You do not need an expert to tell you what Iran is. You need a method to test what the experts are telling you.

This article is part of Why Your Iran Expert Might Be Wrong. For the full data on academic political orientation, see The 8:1 Problem. For the ten structural filters that shape coverage, see Ten Filters.

Footnotes

  1. Thomas Erdbrink, “From Theocracy to ‘Normality,’” The New York Times, February 2019; Al Jazeera, “Iran revokes New York Times correspondent’s accreditation,” June 2019 ↩

  2. Brookings Institution, “Deal or No Deal?” panel transcript, April 2015 (Suzanne Maloney remarks) ↩

  3. Noam Chomsky and Edward Herman, “Manufacturing Consent,” 1988 (comparative framing analysis of KAL 007 vs. Iran Air 655 media coverage) ↩

  4. TIME, “Death Toll in Iran May Already Be in the Thousands,” January 2026; Iran International, independent medical network estimates, January–February 2026 ↩

  5. Reuters, “Special Report: Iran’s crackdown on unrest killed over 1,500,” December 2019; ACLED, “Middle East Overview,” February 2026 ↩