The Bad Proxy: Why Zionist Is Becoming a Civil-Rights Classifier Just as It Stops Measuring Jewish Belief
Universities are asking the word Zionist to do legal work as a signal of identity-based targeting — just as surveys show the label no longer reliably measures the beliefs it is meant to proxy. An investigation into the measurement machinery underneath the fiercest speech debate on campus.
By MyAudioBooks.ai ·
Buried in New York University's student conduct guidance is a sentence that asks one English word to do an extraordinary amount of legal work. The guidance explains that anti-Zionist conduct could trigger the university's obligation to investigate shared-ancestry discrimination — that in certain contexts, the word "Zionist" may function as a code, a stand-in, a way of singling out Jews or Israelis without naming them. The same document is careful about the limits: Zionism itself is not a newly protected characteristic, and criticism of countries or ideologies remains permitted. Read closely, it is a cautious, lawyerly paragraph about context. Read at a distance, it is something else: a major institution announcing that a contested political word has been admitted into the machinery of civil-rights enforcement, with everything that admission implies.
Whether that machinery will work depends on a question almost nobody in the debate is asking, because it sounds like a technicality. The question is not whether anti-Zionism can be antisemitism, or whether criticism of Israel is protected speech. Those questions have been argued to exhaustion. The question is whether the word "Zionist," treated as a signal, still reliably measures what institutions need it to measure — and the most recent survey evidence suggests the answer is deteriorating at exactly the moment the word is being asked to carry more weight.
Start with the numbers, because they are stranger than either side of the argument tends to admit. A national survey of American Jews reported that about thirty-seven percent identify as Zionist — while about eighty-eight percent support Israel's right to exist as a Jewish and democratic state. A Canadian longitudinal study, led by the sociologist Robert Brym, found that forty-nine percent of surveyed Canadian Jews did not identify as Zionists, while ninety-four percent supported Israel's existence as a Jewish state. Among those who declined the label, eighty-eight percent still supported that proposition, and sixty-seven percent reported emotional attachment to Israel. Only about one percent of the full sample identified as anti-Zionist. And in perhaps the most quietly damning finding for anyone hoping to build policy on this word, twenty-nine percent of respondents changed their answer about being Zionist between two survey waves. The label is not merely unpopular with a subset. It is unstable in the hands of the very population it is supposed to describe.
Community-level studies tell the same story at higher resolution. In several Brandeis-affiliated surveys of American Jewish communities, roughly forty percent of Chicago Jews, forty-two percent of Los Angeles Jews, and twenty-six percent of Portland Jews identified with the label — even as much larger majorities in the same communities registered attachment to Israel or support for its existence as a Jewish state. Whatever these people believe, enormous numbers of them will not say it with this word.
A separate Brandeis study helps explain why. When the researcher Matt Boxer asked North American Jews open-ended questions about what Zionism means, the answers did not converge on a definition. They scattered across incompatible narratives: a liberation movement, a religious obligation, a political scandal, an identity, an embarrassment, a duty, a word other people weaponize. Closed survey questions conceal this; they force one check-box onto a term that different respondents are using to mean different things. Statisticians have a phrase for the slow version of this process: semantic drift, the tendency of a word's meaning to migrate across communities and across time until people who appear to be answering the same question are not.
At My Audio Books dot A I, you can create your own audiobooks from prompts, turn your documents into audio, all with one subscription, and store your items in your own personal library.
Now consider what the legal system is trying to do with this migrating word. Title VI of the Civil Rights Act bars discrimination on the basis of race, color, and national origin in programs that receive federal funding. Federal enforcement has long treated Jewish ancestry and Israeli national origin as covered categories — while treating political viewpoint as something else entirely, a domain protected by the First Amendment rather than policed by civil-rights law. That architecture creates a genuine puzzle when conduct is aimed at "Zionists." If the word is functioning in a particular incident as a mechanism for singling out people for their Jewish ancestry or Israeli origin, the conduct may fall inside the statute. If it is functioning as political speech about an ideology, it falls outside. Everything rides on which thing the word is doing, case by case, context by context.
Lawyers have a term for what the word is being asked to be: a proxy, a measurable stand-in for something harder to observe directly. A smoke alarm does not detect fire; it detects smoke, because smoke is the visible trace of the thing that matters. A credit score does not measure character; it measures repayment history, because repayment history is the observable trace of reliability. Institutions reach for proxies whenever the trait they care about — intent, ancestry-based hostility, dangerousness — cannot be read off the surface. And the central fact about every proxy, the fact that decides whether it helps or harms, is that it has error rates in both directions. It can fire when the real thing is absent, and it can stay silent when the real thing is present.
The two error types have names, and they are worth learning because the entire institutional question turns on them. The first is sensitivity: how often a detector catches the cases it is meant to catch. A highly sensitive tripwire for ancestry-based harassment flags nearly every incident where someone is being targeted for being Jewish or Israeli; a less sensitive one lets real cases pass unseen.
The second error runs the other way, and its name is specificity: how often a detector avoids flagging cases it is not meant to catch. A highly specific tripwire stays silent when the speech is genuinely political; a less specific one cries wolf over lawful argument. No real-world classifier scores perfectly on both, and policy chooses, consciously or not, which error it prefers to make.
The campus-climate data lets us see both errors not as abstractions but as identifiable populations. A twenty twenty-four Brandeis study surveyed non-Jewish students at sixty campuses and sorted them by attitudes, and the result was not a spectrum but four distinct clusters. Sixty-six percent expressed hostility to neither Jews nor Israel. Fifteen percent expressed intense hostility toward Israel without holding traditional anti-Jewish beliefs. Sixteen percent held anti-Jewish beliefs without intense hostility toward Israel. And two percent scored high on both. Read that table as an engineer would. A classifier keyed to the word "Zionist" will treat the fifteen percent as its positive cases, though much of that group's speech is political by any ordinary definition — false positives, baked in by design. And it will miss the sixteen percent almost entirely, because people who hold anti-Jewish views do not need the word to act on them; they have older vocabularies, and a tripwire built on this term will never see them — false negatives, also baked in.
This is the measurement problem, stated without any politics attached: the word does not map cleanly onto either the beliefs institutions want to protect students from or the speech the First Amendment requires them to leave alone. It is a noisy channel in both directions. And into this noisy channel, universities are now routing due process, investigations, discipline, and federal funding.
At My Audio Books dot A I, you can listen to this story and thousands of others that explore the hidden science and mechanics behind the headlines.
The legal theory is contested at exactly the point where the measurement is weakest. In a twenty twenty-five Harvard Law Review Forum essay, the scholars Benjamin Eidelson and Deborah Hellman argue that Title VI claims built on anti-Zionism face serious statutory and constitutional hurdles: the statute reaches race, color, and national origin, and a political or religious viewpoint is none of those things, however entangled with identity it may become. On their reading, treating anti-Zionist speech as presumptively discriminatory stretches the statute past its text and collides with the First Amendment's protection of political argument. In a companion response, the legal scholar Stephen Sachs argues for a narrower but real pathway: when institutions or groups require Jews or Israelis to renounce nationhood or ancestral affiliation as a condition of participation — when the price of entry to a club, a classroom, a movement is disavowal — exclusionary anti-Zionism can operate as national-origin discrimination, whatever vocabulary it wears. These are not fringe positions shouting past each other. They are the two serious accounts of where the statute's edge lies, and they disagree precisely about how much evidentiary weight a word can bear.
Notice that the strongest versions of both theories are contextual. Sachs does not claim the word is always discrimination; he claims it becomes discrimination when it is used to condition participation on disavowal. Eidelson and Hellman do not claim anti-Jewish harassment is imaginary; they claim viewpoint-based rules are the wrong instrument for it. NYU's own guidance, read fairly, sits in the same register: the word "could" trigger obligations, in context, when it functions as a mechanism for singling people out — and Zionism is not itself a newly protected class. Harvard's January twenty twenty-five agreement with the Department of Education's Office for Civil Rights covers actual or perceived shared ancestry, explicitly including Palestinian, Arab, Muslim, Jewish, and Israeli identities alike, and likewise does not anoint any ideology as protected. The careful documents all say the same thing: the word is evidence, not verdict.
The enforcement record, though, is where careful doctrine meets institutional incentives, and it must be handled honestly. A twenty twenty-five report from the American Association of University Professors and the Middle East Studies Association examined Title VI complaints filed after October seventh and argued that enforcement frequently treated criticism of Israel or of Zionism as antisemitism without reference to Jews or Judaism at all. The report is an advocacy document with an explicit viewpoint, and it should be read that way — its authors are parties to the argument, not referees. But its complaint dataset is real, and the pattern it documents is exactly the false-positive pathway the survey data predicts: if the word is treated as sufficient evidence, political speech will be reclassified as discrimination at some rate, because the word's political usage is common and growing. You do not have to adopt the report's conclusions to see that its central warning is mathematically coherent.
Meanwhile, the word's definitional collapse continues in the background, indifferent to the legal weight being stacked on it. The same surveys that show low identification show something more subtle: attachment to Israel and support for its existence remain high even among those who refuse the label, which means the label is shedding its descriptive function while retaining its social charge. A word that people will not apply to themselves but will apply to others is a word in the middle of changing jobs. It is becoming less a census category and more an accusation — or, depending on the room, a badge. Either way, it is becoming a worse instrument precisely as institutions calibrate around it.
There is a name for the broader process, and it completes the picture. Juridification is the conversion of a social or political dispute into legal categories and legal procedures — the moment when an argument that used to happen in newspapers and dorm rooms starts happening in complaint forms, investigation protocols, and funding conditions. Juridification is not inherently bad; it is how societies protect people when norms fail. But it has a known side effect: legal categories are sticky. They freeze the meaning of the words they incorporate, because a statute or a guidance document cannot update its vocabulary every survey wave. A political word mid-drift, admitted into a legal instrument, gets embalmed at whatever meaning it held on the day of admission. The institution is then administering a definition the living language has already left.
The scholars of the Nexus Project, in a twenty twenty-five essay titled Governing Jews, made the adjacent argument that legal definitions of antisemitism risk prescribing an official, state-aligned conception of Jewish identity — law deciding not just what discrimination is but what Jewishness means. That essay deserves the credit it has earned, and this article is not repeating it. The claim here is narrower and, frankly, more boring, which is its strength: before law can govern an identity, it must measure something, and the word currently being asked to do the measuring fails basic tests that any engineer would run before trusting a sensor. The two arguments are complements. One is about power; this one is about instrumentation. A proxy this noisy would be rejected in any other enforcement domain — no court would convict on a smoke alarm that fires a third of the time in empty rooms and sleeps through a sixth of real fires.
The campus experience data suggests the cost of mismeasurement is not theoretical, and it does not fall on only one group. Columbia's twenty twenty-five HEALS report, which surveyed more than nine thousand students, found low belonging and high fear around expressing views among both Jewish students and Muslim students — a shared institutional failure that one-group explanations cannot describe. Brandeis's Drawing the Line study of more than two thousand Jewish undergraduates found the same nuance the legal debate keeps flattening: near-universal rejection of claims denying Israel's right to exist, alongside a majority who did not consider criticism of Israel's treatment of Palestinians to be inherently antisemitic. The students closest to the question are drawing finer distinctions than the policies written in their names.
The strongest case against this entire framing deserves to be stated at full strength, because it is genuinely strong. It goes like this: discrimination law has never required proxies to be population-perfect. A code word is evidence in an incident, not a census variable. The question in any real case is what this speaker meant by this word in this context, and for that question, survey stability is beside the point — a word can be abandoning its descriptive function among Jews while retaining perfect clarity in the mouth of someone using it to exclude. NYU's guidance, the counterargument continues, already says exactly this: it requires context, disclaims any new protected class, and permits criticism of countries and ideologies outright. The surveys measure how people label themselves; they say nothing about how antagonists deploy the label. An institution that waited for a statistically pure detector before investigating harassment would never investigate anything.
Every part of that is fair, and the thesis of this article does not survive unless it is conceded. So let it be conceded: the word can function as an identity proxy in a particular incident, contextual judgment is the correct legal standard, and the careful guidance documents already say so. The claim that survives is the institutional one. A standard that works in the courtroom of a single case can still fail in the factory of a bureaucracy. Universities do not process incidents with judges; they process them with administrators, flowcharts, complaint portals, and federal deadlines, and bureaucracies convert every "it depends on context" into a field on a form. The measurement question is not whether context matters — it is whether a contested, drifting word can survive the conversion into form fields without its error rates hardening into false findings against the politically outspoken and missed findings for the actually targeted. The surveys do not prove the policy is wrong. They prove the policy's central variable is unstable at the moment the policy is leaning on it hardest, and that is a thing an institution is entitled to know about itself.
What would change the conclusion? Four findings — four falsifiers, stated as concretely as the claim. First, if incident-level enforcement data showed universities consistently requiring contextual evidence and never treating the label alone as proof of protected-class targeting, the bureaucratic-decay worry would lose its foundation. Second, if survey experiments showed the word remains a reliable and accurate predictor of the relevant identity or belief among the populations these policies actually touch, the measurement premise would collapse outright. Third, if longitudinal data showed institutional protection increasing rather than decreasing the label's stability and people's willingness to use it, the drift argument would have to be abandoned. Fourth, if the gap between self-identification and Israel attachment disappeared after correcting for question wording, age, ideology, or sampling, there would be no measurement story left to tell. None of these studies exists yet, which is itself part of the point: the word was admitted into the machinery before the instrumentation was validated.
There is one more asymmetry worth naming before the ending. Every error this system makes is legible as betrayal to someone. The false positive reads to a student organizer as proof that advocacy is being criminalized. The false negative reads to a Jewish student as proof that the institution cannot see what is happening to her. Both readings will sometimes be correct, because both errors are real, and a policy that can produce both at scale is manufacturing grievance in both directions while calling it protection. The Columbia data — Jewish and Muslim students alike reporting that they do not feel they belong and do not feel safe speaking — is what that manufacturing process looks like from inside.
The open question is not whether universities should protect students from ancestry-based harassment; they must, and the law rightly insists. It is whether protection requires freezing a contested political word into an identity category at the exact historical moment the word is coming loose from any fixed meaning — whether an institution can honor the context-dependence its own lawyers wrote into the guidance while operating at the scale its own bureaucracy demands. The surveys will keep coming, and the number after the word will keep moving, and every policy built on the word will move with it whether it admits that or not. A classifier can drift. A principle should not. The institutions that learn to tell one from the other will protect their students; the ones that cannot will keep prosecuting the wrong word, in both directions, and calling it justice.
At My Audio Books dot A I, you can create fiction, non-fiction, and turn your documents into audio, all stored in one place with a single subscription — plus get instant access to thousands of audiobooks and deep-dive investigations. Learn more today at My Audio Books dot A I.