A live test of Common Ground's core consensus-finding mechanism. Vote Agree, Disagree, or Pass on each statement — or flag when a statement genuinely doesn't reduce to one of those, and say why — no replies, no arguing, just votes, and watch where genuine common ground and real divides actually sit. This pilot's first live topic is National Immigration; more topics open over time (see Topics below).
Most people, across nearly every divide, want to be heard fairly, treated with respect, and to build a life that genuinely feels worth living — and agree with each other far more than it usually looks like, once the loudest and most exaggerated voices stop crowding everyone else out.
This project does three things, in order. First, it finds and makes visible the real agreement that already exists beneath surface-level conflict on a given issue — across cultures, beliefs, and circumstances — and is just as honest about where agreement genuinely isn't possible, rather than paper over real differences with a false consensus. Second, it does the same thing for solutions: anyone can propose a fix, AI brings forward the best peer-reviewed evidence available for and against it — never filtered to favor a popular idea over an unpopular one — and the group is tested, the same rigorous way, for where real agreement on a way forward actually exists. Third, where that tested agreement is real, we work to connect it to an actual path to act on it — an existing government process or a decision-maker who can act — rather than publishing a strong result and hoping someone notices.
We don't start by deciding what counts as progress and asking people to agree with us. No single person, company, or government gets to decide what counts as true or agreed-upon — everything here, including any fix someone proposes, is open to challenge. Everyone's input counts, including people whose own governments would punish them for speaking, and including consideration for those who can't speak for themselves. What we screen out is narrow and specific — threats, harassment, and the like — never a position just because it's unpopular or uncomfortable. The way this works — the method, the code, the reasoning behind it — is open for anyone to inspect; what stays protected is narrower still: the live operational detail a bad-faith actor would need to game it, and the identity of anyone whose safety depends on not being named. We measure success by whether people feel genuinely heard and whether real agreement actually moves somewhere — not by who won an argument. And we say plainly when something here doesn't work yet, or when a better approach comes along — this is a living project, not a finished answer.
Search for a topic, or describe one you're interested in — existing topics are checked first, and only if nothing already covers it does the AI offer to check it as something new. Filter by scope or category to narrow what's live. This pilot currently runs exactly one topic, so most of what's below will look thin until more exist — built to work at that scale, not dressed up to look busier than it is.
Only one topic is live right now, so all three lists above show it — they'll actually differentiate once more topics exist.
How this platform is built to work fairly and resist gaming — a standing feature of every deliberation run here, not a pilot-only extra, meant to stay available to every participant going forward. Not things to vote on, but worth knowing about: how statements get checked, who's answering, and how we guard against gaming.
How it works. Every statement — the ones already here and any you add — is checked against eight known question-wording problems: comparative bundling, vague degree, loaded or presupposition framing, false dichotomies, connotatively loaded wording, double negatives, ambiguous population referents, and modal “should” ambiguity. When you add your own statement, this check runs automatically and in real time before it publishes — if it finds something, it drafts a suggested fix and shows it as the default, with your original one click away. It never blocks you from publishing your original wording.
Why we built it this way. Loosely-worded statements produce misleading data — someone can end up agreeing or disagreeing with a claim for a reason that has nothing to do with the actual disagreement underneath it. The fix is to catch that automatically, at the moment someone writes a statement, rather than after the fact. And it's deliberately narrow: this checks how a statement is worded, never what position it takes — it would flag "immigration is seriously bad" and "immigration is seriously good" identically, for the same vague-degree problem, regardless of which one you agree with.
You can also run this same check yourself on any statement below, including ones you've already voted on, to see exactly what it finds and why. Each check is a live model call and may take a few seconds — you may be asked to allow it the first time.
This topic is scoped nationally, but a fair question is whether the people actually answering are mostly inside the U.S. or mostly from elsewhere — either could shape the result in a way worth knowing about. Rather than a map of individual respondents (easy to fake, and a real re-identification risk at fine grain — see core primitive #28), this shows one aggregate line, self-reported at onboarding, always optional, and only shown once enough people have shared it that no individual could be picked out from the count.
Why behavioral signals, not location. A self-reported or inferred location is exactly what's easiest to fake — the best-documented real case is the FCC's 2017 net neutrality comment docket, where roughly 80% of 22 million comments used forged identity and address data specifically to look like authentic local input. So instead of trusting location, this checks patterns already present in the pilot's own data: many votes landing the same way in a short window, near-identical submitted text from supposedly independent participants, and voting that moves in lockstep across many statements — see core primitive #29.
What a flag does and doesn't do. Nothing here is ever auto-removed, blocked, or reweighted. A real deployment would route a flag to a human curator for judgment; this platform has no separate curator role staffed yet, so the same information is simply shown here directly. The thresholds below are first-pass placeholders sized for a small pilot, not derived from formal research — an open item logged in the project's own docs.
Any passage this project publishes — a statement, a piece of sourced research, this project's own writing about how it works — can be selected the same way you'd select text to copy it. Doing that opens a small menu: ask AI to explain the passage in plain language, suggest clearer wording (shown to everyone reading that passage, like a wiki edit suggestion — never applied automatically), flag it as unclear with one tap and no writing required (this becomes a signal for whoever curates this pilot, not a public comment), or ask an open, free-form question about just that passage. On sourced research and legal passages, a fifth option lets you challenge a specific fact directly — see the panel below for how that works. Nothing here changes what you're reading: the original text always stays exactly as it was, and everything you do here creates a new, separate, dated entry alongside it.
Every piece of research this project publishes — a statement's sourced facts, a topic's legal or background detail — can be wrong, incomplete, or go stale, the same way any other research can. Anyone can point at a specific passage they think is mistaken or misleading, explain why, and point to where the correcting evidence can be found. AI checks the challenge the same way a professional fact-checker would — comparing it against independent, established sources, not just reading deeper into whatever was cited — and rates it on a graded scale: confirmed and well-sourced, true but missing important context, genuinely disputed among legitimate sources, or not supported. Nothing here gets silently rewritten: the original text stays exactly as it was, and a dated, sourced correction is attached directly at the passage in question, the same way a newspaper runs a correction rather than quietly editing yesterday's article. As of today, this check is done by AI without a separate human curator role staffed to review it — the same honest gap this project already names for its other flagging mechanisms — and it can only work from evidence that's linked to (a page, a document, a citation), not a file uploaded directly, which this platform doesn't yet support.
Whenever someone flags a highlighted passage as unclear (see "Highlighting text to ask for help" above), it's logged here — never publicly, and never as a comment on the passage itself — so whoever curates this pilot can see which of its own passages keep tripping people up, and why, the same closing-the-loop idea behind "it depends" on statements above.
Some facts this project publishes age fast; most don't. Who currently holds a named office or committee seat, and a bill's current legislative status, can change on the real world's own schedule — a Senator's roster spot can change from one day to the next. Historical research, survey findings, and constitutional or statutory text barely change at all. Treating everything as equally permanent is what let a since-deceased senator's name sit on this pilot's own committee roster until it was caught directly (see the Corrections & Challenges log above for that fix) — so only the genuinely fast-changing content gets the standing process below; nothing else needed it.
What's checked, and how often. The committee roster and this topic's three tracked bills are re-checked once a month against their own official sources — the Senate Judiciary Committee's own membership page, and congress.gov for each bill. A confirmed change publishes immediately, tagged and logged the same way an AI-verified correction is (primitive #35) — no waiting on a review step, since this reuses the same source-verification standard already trusted elsewhere in this project. Every monthly check leaves a dated record, including a check that finds nothing has changed, so there's always a real answer to "when was this last verified," not just a claim.
What this doesn't cover yet. This starts narrow — the roster and these three bills only — and is planned to widen to every fast-changing fact across the whole project before this goes live to real public participants, alongside finishing this topic's legal-precedent research (primitive #34) across all its tracks, not just the three that currently have it. Both are tracked, decided commitments, not yet built.
The merge-suggestion half of this is live — see the "Topics" panel above, where a real AI check tells you whether your requested topic is already covered before anything new gets created. This pilot still only has one topic actually running, though, so what's below stays a worked example of the fuller scope-advisory conversation (residency and thin-local-data disclosures) that would happen once a genuinely new, locally-scoped topic could actually go live — using the same never-blocks, AI-suggests / human-decides pattern used for statement wording checks above. Nothing below is wired to anything live.
Whichever the person picks, the same two facts — no residency check, and thin local data — would also show up in the topic's own public brief, not just in this one private exchange with whoever proposed it.
Designed, discussed, and logged in the project's docs — not active here yet. Shown so the shape of where this is going is visible, not just described.
A topic's subject doesn't change, but what people think about it can — this pilot now keeps an append-only record of every vote, including any later change, rather than only your latest answer. See "Your view over time" (your own private record) and "How opinion is shifting over time" (the aggregate, cohort-gated picture) further down this page — core primitive #25.
An optional prompt after you vote — "what's shaped your view on this?" — for a citation, documentary, or claim behind your answer. Checked for accuracy, never used to correct or argue with you individually; only ever surfaced as an aggregate pattern or a new educational note.
Every new statement submitted through "Add your own statement" now gets a real, live AI check automatically — if it finds a real problem (bundling, vague degree, loaded framing, and more), it drafts a suggested fix and shows it as the default, with your original one click away. The quick pattern-match hint above the box (try typing "bigger... than" or "seriously") still runs as you type, as an early nudge before the real check runs on submit. See "How statements get checked for fair wording" further down this page to run the same check yourself on any existing statement, and to read why it's built this way.
A way to attach the personal or cultural story behind a view — not as a footnote, but able to move to the center of a discussion when that's genuinely what the disagreement turns on.
Plain-language, verified explainers for statements people flag as "I'd need more facts to judge this fairly" — full technical depth available on request, but never required to participate.
Ten of the statements above (tagged “AI-researched”) come from historical/causal research into why immigration conflict recurs, organized into six independent tracks — economic, cultural, social-cohesion, fiscal, sovereignty vs. humanitarian authority, and historical instrumentalization by power — rather than one curator's first draft. The rest of the statement set was rebuilt and organized into five further tracks the same day (rule of law & enforcement, sanctuary & local enforcement, legal pathways & policy, climate-driven migration, and one untracked statement about discourse quality), so the whole set now reads as one coherent map rather than two separately-added batches. There's no separate “range” control anymore — a topic's scope comes from how it's named (this one: “National Immigration”), not a dial.
The "Background" note above the statements is a first, sourced pass presenting factual perspectives on the issue's history — not framed as "two sides," since this rarely reduces to one axis, and not claimed as every perspective there is, since a first pass inevitably misses some. It hasn't yet been reviewed by people who'd frame this history differently, which is the standard this project holds itself to before calling a brief like this settled.
A button beside every statement — see it above under "It depends — tell us why" — opens the sourced research behind that statement's own track, entirely optional and never shown automatically. Research is attached at the track level (11 tracks' worth, extended 2026-09-06 to cover the five newer curator-organized tracks alongside the original six), shown from multiple angles wherever more than one genuinely exists, and framed the same regardless of how you vote — never as a reply to your answer.
Inside "Explore the Research," on the three tracks where it's most directly relevant (Rule of law & enforcement, Sanctuary & local enforcement, Legal pathways & policy), a distinctly labeled section now shows the actual governing law and controlling court precedent behind that track — never whether a statement itself "is legal," only what a real, cited case has actually held, and whether that area of law is currently settled or still being fought over in court. General legal information, not legal advice.
There's no separate "range" control anymore — scope comes from how a topic itself is phrased ("immigration in Terlingua, TX" instead of a dial). Before a locally-scoped topic like that would go live, the same AI check already used for statement wording (see "How statements get checked for fair wording" further down) would run on the topic's framing too — telling whoever proposed it what they will and won't get, offering a better-fitting alternative, and letting them keep their original framing anyway. A worked example of that conversation is below.
Every proposal submitted through Solutions above now gets an automatic AI check — but only for conduct (threats, harassment, doxxing, targeted slurs, spam), never for tone, emotional intensity, or the position taken. Research on "civility"-style content moderation finds it disproportionately silences legitimate speech — hate-speech classifiers misclassify African American English as offensive at roughly double the normal rate, and studies of human moderators find similar tone-policing effects — so this project screens conduct only, and never silently removes a flagged item; it stays visible with a "flagged for review" note instead. Who actually reviews a flagged item is still an open staffing question, same as vTaiwan's own unstaffed facilitation stage above. This screening — like the rest of "Under the hood" — is meant to run standing behind every topic on this platform, not just this immigration pilot.
The AI-drafted synthesis card that appears automatically in Solutions above (2026-09-07: no button anymore — it drafts on its own once there are at least two proposals with votes, and is added inline, clearly labeled "✨ AI-drafted synthesis") is a first, simplified build of a real published mechanism — Google DeepMind's "Habermas Machine" (Science, 2024), which drafts a candidate statement designed to earn the highest agreement across a group's differing views. In DeepMind's own testing, groups preferred its drafts over human mediators' 56% of the time and were measurably less divided afterward. This build does one drafting pass from current proposals and votes, not the full multi-round critique-and-refine loop DeepMind tested — and, as DeepMind's own researchers note of the underlying method, it doesn't fact-check or moderate a discussion on its own, which is why it's paired with the evidence and conduct-screening steps above rather than standing alone.
The Solutions set above (core primitive #30) was broadened using real public-arena research rather than one curator's own sense of what matters — Migration Policy Institute's own named issue taxonomy checked against Pew/Gallup/PRRI polling data, covering interior enforcement/ICE (split into three separately-votable pieces: accountability reform, enforcement-priority scope, and funding/staffing — the piece Michael specifically named as a major issue, confirmed rather than assumed), border security, asylum & refugee policy, birthright citizenship (shown as a genuine, paired either-way choice rather than one side), guest-worker visas, family sponsorship, and DACA/Dreamers. To avoid overwhelming anyone with all of it at once, a handful with the strongest salience grounding show by default, with the rest reachable via "See more solutions" — reusing the same layered-depth pattern (primitive #10) already used elsewhere in this pilot, not a new mechanism.
"Who's answering, in aggregate" above (core primitive #28) shows one self-reported, optional, threshold-gated line — the rough share of respondents inside vs. outside the U.S. — rather than a map of individual locations, which real research shows is both easy to fake (the FCC's 2017 comment-fraud precedent) and a re-identification risk at fine grain. Never used to weight, verify, or gate a vote.
"How we guard against gaming the results" above (core primitive #29) checks for the behavioral signals real research says actually catch coordinated or inauthentic participation — timing clusters, near-duplicate submitted text, lockstep voting — rather than trusting a location claim, which is exactly what's easiest to fake. Flags a pattern for human judgment; never auto-removes, blocks, or reweights anything on its own.
A tested, cross-cluster-agreed solution is meant to go somewhere, not just be published — aimed at a real, existing decision point wherever one exists (a rulemaking docket, a ballot-initiative process, a local referendum), or otherwise routed through a mandatory human-facilitation step plus a permanent public record of who was solicited and what they did with it, modeled on vTaiwan and on what worked (and didn't) in Ireland's citizens' assemblies. Not built into this pilot yet — there's no real decision-maker on the other end of this specific test.
This round is scoped nationally, not to one town or region — its purpose is not to settle who's right on immigration, but to map where people already agree, clearly and specifically, and where real disagreement remains, so those areas get named honestly instead of assumed.
Research on immigration conflict finds several separate causes — not one root cause. Someone can worry that immigration strains social cohesion while also believing it enriches culture; those two beliefs don't move together, which is why this rarely comes down to one single disagreement. Decades of research find something similar about competition: when people believe they're competing with another group for jobs or status, that belief alone creates friction — whether or not the competition is actually real. Researchers call this "Realistic Group Conflict Theory." Migration Policy Centre · Realistic Group Conflict Theory
The economic and fiscal effects are real, but they land unevenly. Over a decade, immigration's overall effect on native-born wages is small — but that effect falls hardest on workers without a high school diploma, and on immigrants who arrived earlier, since they compete most directly for the same jobs. On the fiscal side, immigrant families cost local governments more than they pay in during their first few years here — but the same research finds a net benefit to federal finances over 75 years, with the U.S.-born children of immigrants paying in the most of all. National Academies of Sciences (2016)
Some of this isn't a factual dispute at all — it's a genuine disagreement about values, and no amount of evidence resolves it. Put plainly: does a country get to set its own rules for who may enter, full stop — or does someone fleeing real danger have a claim that comes first, regardless of what a country has already decided? People land on real, different answers to that question, not because one side has the facts wrong. Harvard Edmond & Lily Safra Center for Ethics. History adds an uncomfortable wrinkle: immigration limits haven't always been about economics or culture — sometimes they've been a tool for consolidating political power. The 1924 U.S. Immigration Act, for example, set strict quotas built on the racial-hierarchy thinking of the eugenics movement, not on economic concern. Public Books
LegalCongress and the executive branch hold broad, historically court-deferred authority over immigration — courts call this the “plenary power” doctrine, and it's a real part of why immigration law works differently than most other areas of law. A separate, well-settled rule limits how far that federal authority reaches into local government: the Supreme Court has held that Washington can't force state or local officials to run a federal program, which is the actual legal basis most “sanctuary” policies rely on. Both doctrines remain controlling law as of September 2026, tested directly against real cases — Arizona v. United States (2012) for the first, Printz v. United States (1997) for the second.
Not every corner of immigration law is this settled, though. DACA — the executive program shielding people brought to the U.S. as children from removal — has never been given permanent statutory protection by Congress, and the Supreme Court's 2020 ruling on it decided only a procedural question, explicitly declining to rule on whether the program itself is lawful. And a genuinely live dispute is playing out in court right now over birthright citizenship itself: the Supreme Court reaffirmed it in June 2026, new executive orders tried to narrow it five weeks later, and as of this writing that fight remains unresolved. This is general legal information, not legal advice.
This is a first, sourced pass at the history — not yet reviewed by people who'd frame it differently, so treat it as informative, not final. Every statement below is organized into one of eleven tracks; look for the small tag above each one (a colored "AI-researched" tag for the six tracks that came directly from this research, a plainer tag for tracks organized by curation).
This is a small, informal pilot among people invited to this link — not a public poll. You can join with just a nickname or stay fully anonymous; nothing you type here is checked against a real identity.
Welcome back — your name and region are saved on this device from last time.
One at a time. Vote however you actually feel — there's no wrong answer, and passing is a real option. If a statement genuinely doesn't work as agree/disagree, say so — that's real signal too, not a cop-out. Every statement belongs to one of eleven tracks (economic, cultural, social cohesion, fiscal, sovereignty vs. humanitarian claim, historical instrumentalization, rule of law & enforcement, sanctuary & local enforcement, legal pathways & policy, climate-driven migration, or untracked) rather than one single framing — statements tagged “AI-researched” came directly from the project's historical/causal research (see Background above); the rest were organized into their track by curation.
Every statement now gets a real, live AI check for bundling or vague wording before it publishes — if it finds something, you'll see a suggested fix and can use it or keep your original, one click either way.
Solutions for this topic — vote on them the same way as any statement. Every proposal gets an automatic AI check: for conduct only, never tone or the position taken, and for evidence that could support it or complicate it. The starting set below is broader than one curator's first draft — it's checked against real research on which solutions actually have public currency (Migration Policy Institute's own named issue areas, Pew/Gallup/PRRI polling), so it doesn't just reflect one person's sense of what matters. To keep the page from feeling like a wall of options, a handful with the strongest research grounding are shown first; the rest is one click away under "See more solutions" below, not left out. Once there are at least two proposals with votes, the AI also drafts its own synthesis of where they overlap automatically — no button to press — and adds it above as its own card, clearly tagged "✨ AI-drafted synthesis," never swapped in for the proposals it's drawn from. See "What's coming to this pilot" for how this set was built.
A solution isn't one atomic claim, so it's broken into four parts: a mechanism, who it's meant to help, who may bear a cost, and what's genuinely uncertain. Once you submit, the same automatic AI checks run on your idea as on everyone else's, and it's added above as its own card for people to vote on.
Agreement on its own has a well-documented failure mode: France's Citizens' Convention on Climate saw roughly 60% of its 149 proposals dropped or weakened even after "unfiltered" was promised, and separate research (Gilens & Page, 2014) found that across 1,779 real U.S. policy questions, ordinary citizens' preferences had close to no independent effect on what became policy once economic elites' preferences were accounted for. Genuine, well-run consensus is necessary — it just isn't sufficient on its own. So here is the honest, currently-conceptual answer to "then what": how a tested, cross-cluster-agreed statement or solution from this topic could actually reach someone with real decision authority, not just get published and hoped for.
When a real door already exists — a court requirement, a rulemaking docket — route a tested result there instead of just publishing it and hoping.
Ireland's citizens' assemblies are the clearest natural experiment on this: recommendations on marriage equality (2015), abortion (2018), and blasphemy removal (2018) all passed into law — not because Irish deliberation was uniquely persuasive, but because Ireland's constitution legally requires a referendum on any constitutional change, giving those specific recommendations a mandatory door to walk through. The same Convention's other recommendations, with no equivalent legal trigger, were mostly shelved. For immigration specifically, that means routing a tested result toward whatever real mechanism actually exists at the relevant scope — a federal rulemaking docket, a ballot initiative, a local referendum — rather than publishing a strong recommendation and hoping it gets noticed.
As of today, DHS has a genuinely open public-comment period on a proposed $103,265 fee on new H‑1B cap-subject petitions — Federal Register, Docket USCIS-2026-0298, RIN 1615-AD20 — with comments open through September 24, 2026. That's exactly the kind of real door this item describes: a tested, cross-cluster result from this topic's economic-competition and legal-pathways statements (wages, jobs, and skilled-worker policy) would have somewhere real to actually be routed, right now, if this pilot ever reached that stage.
When no such door exists yet, a mandatory human handoff to real officials, and a public record of exactly who was asked and what they did.
Modeled on vTaiwan's own process, which pairs each consensus result with a mandatory human-facilitation stage connecting it to relevant officials — and a permanent, attributable public record of exactly who was solicited and what they did with it, so silence itself becomes visible rather than invisible.
This is the actual shape the tracker would take, built now rather than mocked up — honestly empty, because nothing has happened yet to record:
| Statement or solution | Recipient | Date solicited | Status |
|---|---|---|---|
| No solicitations have been sent yet. This pilot hasn't reached the participation scale where reaching out to a real official or agency would be a meaningful next step. Once it does, every row here will show exactly who was asked, when, and what they did with it — including silence, following the same rule this project already applies to any other party (core primitive #3). | |||
This connects consensus to power — it never tells anyone which direction is the right one.
This stays strictly about making genuine, well-tested consensus land somewhere real — not about "st" or its designers quietly picking which outcome is right. Concentrated wealth and power can still simply reject a solution regardless of consensus quality; that's a real, named limit this roadmap doesn't claim to solve, only to make a little more visible and a little more costly to ignore.
An accountable official's own disclosed position, shown side by side with where their constituents actually stand — never blended, never scored.
For a public official, agency, or company already accountable to the public in some way, their own disclosed position on a given alignment point — a vote, filed legislation, a public comment — shown side by side with, never blended into, where their own constituents or customers actually stand. Deliberately not held to participant-level privacy (core primitives #11/#12): an elected official's votes and a regulated company's public record are already public by design, and the asymmetry is the point, not an oversight. Modeled on real precedent — Project Vote Smart's Political Courage Test, InfluenceMap/LobbyMap, and the CPA-Zicklin Index — including that precedent's own honest warning sign: Vote Smart's participation rate fell from 72% to 20% once candidates realized answers were being mined for attack ads. Strictly descriptive, never a score or a "hypocrisy" verdict — see core primitive #33 for the full design, and why this piece "won't gain traction until, and if this project gets legs."
Statement: “Increased immigration does not measurably affect wages for native-born workers who have a four-year college degree.”
1 vote recorded so far (agree). Genuinely too little data to show a meaningful bridge/divide split yet — shown as 1, not scaled up or hidden.
Sens. Chuck Grassley (R-IA) and Dick Durbin (D-IL), S.2928, the H‑1B and L‑1 Visa Reform Act of 2025 (introduced Sept. 29, 2025): “Major companies are laying off thousands of American workers while filing thousands of visa petitions for foreign workers at depressed wages and poor working conditions.” — Sen. Durbin, in the bill's own introduction announcement.
The full committee that holds jurisdiction over immigration legislation, the real bills currently on the table, and how to reach them — plus, separately, the vote each of them actually cast.
Federal immigration legislation has to clear one committee first: the Senate Judiciary Committee, 22 members, plus its Subcommittee on Border Security & Immigration. Below: what the committee's own chair and ranking members say in public about one real vote, in their own words; the full 22‑member roster and how to reach them; the actual bills currently on the table, each briefly summarized with a link to the real bill text; and, separately, exactly how every one of the 22 voted on the two of those bills that have already reached a floor vote.
“Laken Riley's life was robbed by an illegal immigrant, and repeat criminal offender, who the Biden administration allowed to slip through the cracks. The Laken Riley Act would ensure criminals who have illegally entered our country receive due justice and would help keep communities safe. We must act now to ensure this terrible tragedy never happens again.”— Official statement, Jan. 2025
“The loss of a child is something no parent should have to endure, and we should do everything possible to make sure what happened to Laken Riley never happens again. But… this bill will not accomplish its stated goal. I'm genuinely disappointed in the passage of this bill as it stands and deeply concerned about how it will be implemented.”— Official statement, Jan. 20, 2025
“Laken Riley's death will not be forgotten, and I'm grateful President Trump signed this historic bill into law to honor her legacy and prioritize making our nation safe again.”— Official statement, 2025
“This bill puts a target on the backs of millions of people, including children, requiring mandatory detention of immigrants who are arrested even for minor, non-violent offenses.”— Official statement, Jan. 2025
All 22 members whose votes actually decide what immigration legislation looks like — the people below, not whoever is loudest about the topic elsewhere.
✓ Roster current as of Sep 10, 2026 — checked directly against the committee's own official page, then re-checked automatically on a standing monthly schedule going forward (see "Data currency" under Under the hood).
The four above link directly to their own contact pages. To reach any of the other 22, the Senate's own directory — senate.gov/senators/senators-contact.htm — links to every senator's official contact form; look up their name there rather than guessing a web address.
Roster update, 2026-09-10: Sen. Lindsey Graham (R‑SC) died on July 12, 2026. South Carolina's governor appointed his sister, Sen. Darline Graham, to complete his term, and the full Senate confirmed her committee assignments — including this seat — on July 21, 2026. The roster above reflects that change; the vote table below is a historical record of who actually cast each vote at the time, so it still shows Sen. Lindsey Graham for the two bills voted on before his death. Source: Senate Judiciary Committee — Members.
✓ Bill statuses current as of Sep 10, 2026, checked directly against congress.gov — re-checked automatically each month going forward.
Requires DHS to detain undocumented immigrants who are arrested, charged, or convicted of theft, burglary, larceny, or assaulting a law-enforcement officer, and lets state attorneys general sue the federal government over certain immigration-enforcement decisions.
Read the actual bill on congress.gov →A $70 billion budget-reconciliation package funding ICE and Customs and Border Protection hiring, detention capacity, and border technology through 2029.
Read the actual bill on congress.gov →Would tighten wage, recruitment, and “specialty occupation” standards for employers sponsoring H‑1B and L‑1 visas, and prioritize H‑1B visas for advanced-degree STEM applicants. Introduced Sept. 29, 2025; still in committee, so there is no vote yet to show.
Read the actual bill on congress.gov →| Legislator | S.5, Laken Riley Act | S.2, Secure America Act |
|---|---|---|
| Sen. Chuck Grassley (R‑IA) | ✔ Yes | ✔ Yes |
| Sen. John Cornyn (R‑TX) | ✔ Yes | ✔ Yes |
| Sen. Lindsey Graham (R‑SC) | ✔ Yes | ✔ Yes |
| Sen. Mike Lee (R‑UT) | ✔ Yes | ✔ Yes |
| Sen. Ted Cruz (R‑TX) | ✔ Yes | ✔ Yes |
| Sen. Josh Hawley (R‑MO) | ✔ Yes | ✔ Yes |
| Sen. Thom Tillis (R‑NC) | ✔ Yes | ✔ Yes |
| Sen. John Kennedy (R‑LA) | ✔ Yes | ✔ Yes |
| Sen. Marsha Blackburn (R‑TN) | ✔ Yes | ✔ Yes |
| Sen. Eric Schmitt (R‑MO) | ✔ Yes | ✔ Yes |
| Sen. Katie Britt (R‑AL) | ✔ Yes | ✔ Yes |
| Sen. Ashley Moody (R‑FL) | ✔ Yes | ✔ Yes |
| Sen. Dick Durbin (D‑IL) | ✘ No | ✘ No |
| Sen. Alex Padilla (D‑CA) | ✘ No | ✘ No |
| Sen. Sheldon Whitehouse (D‑RI) | ✘ No | ✘ No |
| Sen. Amy Klobuchar (D‑MN) | ✘ No | ✘ No |
| Sen. Chris Coons (D‑DE) | ✘ No | ✘ No |
| Sen. Richard Blumenthal (D‑CT) | ✘ No | ✘ No |
| Sen. Mazie Hirono (D‑HI) | ✘ No | ✘ No |
| Sen. Cory Booker (D‑NJ) | ✘ No | ✘ No |
| Sen. Peter Welch (D‑VT) | ✘ No | ✘ No |
| Sen. Adam Schiff (D‑CA) | ✘ No | ✘ No |
S.5 final passage: Jan. 20, 2025, 64 Yes–35 No. S.2 final passage: June 5, 2026, 52 Yes–47 No (Sen. Lisa Murkowski, R‑AK — not a Judiciary Committee member — was the only Republican to vote no on S.2; every committee member's vote above follows their full party's reported total, cross-checked against senators' own statements where available). Recorded votes, not this pilot's data — verify any name directly via GovTrack (S.5) or GovTrack (S.2). Sen. Lindsey Graham, shown above, cast both of these votes himself before his death on July 12, 2026; his seat is now held by Sen. Darline Graham (see the roster note above), who was not yet in office for either vote — nothing is shown for her on these two bills, rather than guessing how she'd have voted.
Two real votes on two real bills, shown plainly — not a running scorecard across every issue, and not something this pilot generates a rating from. S.2928 has no vote yet, so none is shown or guessed for it. Every quote, bill link, and vote above traces to its own primary source; no score, ranking, or “hypocrisy” label is calculated from the comparison (core primitive #33).
An AI-synthesized letter drafts itself automatically from this topic's actual vote tallies — no statement to hand-pick. Pick your state, edit every word, then send it yourself.
This can't submit anything to a Senate office on your behalf — almost none accept an automated submission, and doing so would misrepresent who's actually speaking. What it can honestly do: once enough statements have real votes behind them, an AI drafts one letter body synthesizing where this topic's participants actually found common ground and where they were genuinely divided — from the real tallies, not a hand-picked single statement — and you edit or rewrite any of it before copying it to your clipboard and opening your Senator's own official contact page to paste it in and send it yourself. Your state, chosen below, is used only to look up your two Senators for this one tool — it's kept on this device, separate from the "inside/outside the US" line from "Before you start," and is never shown to anyone else or tied to your votes.
The AI drafting this is instructed to state real vote counts plainly and never claim a "consensus" the numbers don't support, never take its own position on the underlying question, and never invent facts beyond this topic's real tallies. It's a shared starting draft, regenerated automatically as more votes come in — a draft to react to, not a form to rubber-stamp, so read it and make it actually yours before it goes anywhere.
Nothing above is wired to a real decision-maker yet — there's no live response mechanism on the other end of this pilot's specific results. This section says so plainly rather than implying otherwise. See "What's coming to this pilot" on the Home page for the rest of what's designed but not active.
Each dot is one participant. Distance between dots reflects how similarly two people have voted across all statements — closer together means more agreement, farther apart means more disagreement. There's no unit or fixed scale on either axis and no "good" direction to be positioned in; only relative closeness carries meaning.
Every vote — a first one and any later change — is quietly logged (never overwritten) so this can show whether views on a statement have genuinely moved as the conversation continues, the same before/after idea behind Fishkin's deliberative-polling research. This is meant to be an active, ongoing output of the project, not a one-time snapshot — check back as more people take part and views develop. Shown only once enough distinct people have a recorded reading on a statement that no one person's own change could be picked out of it.
A topic like this doesn't change, but what any of us thinks about it can — new events, new facts, more time to sit with a question. Every time you change an answer (using "‹ Back to previous answer," any time — today or after coming back to this same link weeks from now) it's added to a private record only you can see here, never shown to anyone else, never tied to your votes in any public view. See core primitive #25.
Statements that hold up across every opinion group, not just one.
Statements the groups read very differently — exactly what's worth understanding, not smoothing over.
Statements enough people flagged as "it depends" that the statement — not the opinion — is probably the thing to fix.
This is the fuller research behind the background brief on the topic page — the same six tracks that produced it, each with its own sourced facts. Nothing here is new; the brief is a short, readable summary of exactly this.
Immigration's effect on wages isn't a flat yes or no — it depends heavily on which workers you're asking about.
These two statements sit side by side on purpose. Research finds people can believe both things at once, or just one — they aren't opposite ends of a single scale.
This worry is tracked on its own, apart from the economic and cultural-identity statements, because research treats it as its own distinct driver — not just another way of naming one of the other two.
The short-term and long-term costs are both real — they just answer different questions, and they come from the same research.
Both findings — costs more locally in the short term, contributes more federally in the long term — come from the same research. Nothing here tells you which one should matter more to you.
This isn't a factual argument to settle — it's a genuine disagreement about who should have the final say. Here are the two ways people frame that question; neither is presented as the right answer.
Both are genuinely held views, not a claim and a rebuttal — nothing here says which one should win.
This is a specific, checkable historical claim about one past use of immigration law — not a claim about why anyone today holds the views they hold.
This is a first, sourced research pass — not a complete literature review, and not yet reviewed by people who'd frame this history differently. It's treated as a standing, expanding body of research, not a finished one.
This describes what courts have actually held, and how settled or contested each area of law currently is — never whether any statement on the topic page “is legal.” General legal information, not legal advice.
The case-law summaries below describe what courts have done with these words. This is the underlying text those rulings interpret — each entry linked to an official, full-text government source rather than a secondary summary.
This is general legal information, not legal advice — it describes what courts and the government have actually said, never whether any statement above "is legal." This covers the three tracks with researched case law so far (rule of law & enforcement, sanctuary & local enforcement, and legal pathways & policy) — not all eleven tracks. Current as of September 2026; law in this area, especially the birthright-citizenship dispute, can and does change.
Not tied to one specific statement on this topic today, but part of the same broader legal picture around enforcement.
Every holding above was cross-checked against at least two independent sources before being included here — this is a curator-level first pass, not an independently reviewed legal analysis, and current only as of September 2026. Law in this area, especially the birthright-citizenship dispute, can and does change.