Each type of information has a label so facts, candidate words, and our analysis are not mistaken for one another.
01
Verified fact
A claim confirmed by a reliable, publicly available source.
02
Public record
A document, decision, or dataset from a public body.
03
Candidate statement
A candidate’s own words, clearly attributed and not presented as fact.
04
Public discussion
A limited sample of accessible public conversation; it does not represent all voters.
05
IntelliSync perspective
Our editorial interpretation or recommendation, clearly identified.
06
AI analysis
AI-assisted analysis that must be reviewed by a person before publication.
How we measure website use
We use first-party analytics to count page views and bounded civic actions such as opening a source, reading the methodology, or successfully recording a poll response. We never store the poll choice, the identity of a candidate viewed, a raw IP address, exact location, browser fingerprint, or free-form search terms. The pseudonymous identifier changes daily, raw events are deleted after 30 days, and all reporting stays inside the protected administrator dashboard.
AI-generated editorial illustration · CK Signals
AI and editorial review
AI and editorial process
From public source to published work.
CK Signals uses AI to help organize public information, not to decide which candidate is better. AI outputs are proposals that remain subject to evidence and human review before anything is published.
01
What AI helps with
AI may help organize large volumes of public information, classify material by provenance, identify topics, apply descriptive tags, summarize documents, extract stated commitments, and propose questions or criterion-level classifications for review.
02
What AI never decides
AI does not decide candidate quality, endorsements, rankings, overall scores, election predictions, or publication. It also cannot automatically turn a statement into a verified fact.
03
How sources are selected
We review public records, verified candidate-owned sources, accessible and inspectable reporting, and accessible public discussion. We check each source’s identity, date, accessibility, and relevance. Private groups, unverified profiles, search snippets, and inaccessible claims are excluded.
04
How the same rubric is applied
For mayoral and ward platform columns, every commitment is examined against the same seven criteria and the same seven-day source window. The authority test changes with the office: it separates what the candidate can initiate from what Council, administration and partners control. Only campaign-owned statements can satisfy a criterion; public records provide context. “Not stated” means no qualifying element was found in that window, not a negative grade or zero.
05
What a person validates
A human reviewer remains accountable for the final framing, source selection and attribution, whether evidence supports each claim, every classification, meaning parity between languages, published limitations, and the publication decision.
06
How uncertainty remains visible
Conflicting evidence, inaccessible information, source-window limits, and unanswered questions remain visible. AI never silently fills an evidence gap and is never treated as a source.
07
How to challenge a result
Anyone can challenge a source, attribution, translation, or application of our rules by email. The request is reviewed against the same evidence standard, and any material correction publicly explains what changed and why.
Direct answers and evidence
Direct answers
What our method means in practice.
How does Chatham-Kent Signals separate facts from opinion?
Every item carries a visible label. Verified facts, public records, candidate statements, public discussion, the IntelliSync view, and AI analysis are never presented as one category.
Do the community polls measure public opinion?
No. They are voluntary, non-scientific, and participant-only. They do not measure voter intention, popularity, or a likely election result.
How are candidates treated fairly?
Every registered candidate receives the same response opportunity and verification path. A missing response is never interpreted as a negative result.
Our public promise
Our public promise
What you can expect from us.
01
We show where information comes from.
02
We separate facts, statements, and our own perspective.
03
We give every registered candidate the same opportunity to respond.
04
When evidence is not strong enough, we say so clearly.
Leadership attribution
Leadership attribution
How we assess accomplishment claims.
An outcome occurring during a term does not automatically become the officeholder’s personal result. The same rule applies to credit and blame.
01
Activity
What was done: meetings, calls, visits, approvals, or spending.
02
Output
What the activity directly produced: spaces, services, cases processed, or projects completed.
03
Programme outcome
The measured change among the people, services, or assets reached by the intervention.
04
System outcome
The municipality-wide trend showing whether the intervention is changing conditions at scale.
Direct-credit test
We require a traceable action or decision, a plausible mechanism connecting it to the result, and an observable measure. NVI means only that no verifiable indexed item was found in the review window; it never proves inactivity.
Version municipal-candidate-commitment-v2 applies the same seven dimensions to mayoral and council candidates. The authority test reflects the office being sought. It does not measure popularity, voting intention or likelihood of winning.
Equal source window
For every candidate, the same seven-day window covers the official registry, every public page listed in a verified campaign-owned site's sitemap, linked documents, and verified public campaign accounts accessible to the review tools. Operational pages and repeated themes are logged without inflating the commitment count. Private groups, unverified profiles, search snippets and inaccessible claims are excluded and the limitation is published.
For a council candidate, ward relevance is considered alongside municipality-wide obligations and trade-offs. One councillor can propose, question, scrutinize and build support; Council collectively decides bylaws, spending, land and policy.
01
Problem and baseline
Does the commitment define the local problem with current, source-verifiable baseline information?
02
Action and intended result
Does it connect a concrete municipal action to an observable result?
03
Authority and ownership
Does it distinguish what the candidate can initiate from what Council, administration and partners control?
04
Timeline and milestones
Does it state commitment-specific deadlines or sequenced milestones?
05
Funding and resources
Does it address costs, funding, procurement, staffing, land or partner contributions?
06
Outcome measurement
Does it identify a baseline, target, outcome measure and reporting checkpoint?
07
Dependencies and decision rule
Does it identify material constraints, safeguards, trade-offs and a scale, change or stop rule?
Documented
All required elements are explicit and commitment-specific.
Partly documented
Meaningful elements appear, but material details remain unstated.
Not stated
No qualifying element was found in the reviewed source window.
Not applicable
Used only when a public explanation shows why the criterion does not apply.
Comparison rule
We may publish descriptive counts, such as how many commitments document a timeline. We never total criteria, calculate a candidate percentage or rank platforms. Missing material becomes INSUFFICIENT EVIDENCE, never zero.
AI-generated editorial illustration · CK Signals
Candidate responses and community polls
Fair participation
How responses are handled.
Candidate responses
Every certified mayoral and council candidate receives the same questions and the same opportunity to respond by email; no candidate portal is required. We verify the sender, preserve the original language, identify the reviewed translation, and ask the candidate to confirm any translation or material edit. Only responses with confirmed source, translation and candidate approval become public. A missing response is never treated as a negative conclusion.
Community polls
Each browser can respond once to each poll. For an active poll, this browser sees the breakdown only after responding and after the privacy threshold is met. Polls are voluntary and non-scientific; their results do not represent Chatham-Kent as a whole.
Public discussion
We use only publicly accessible discussion. We do not publish candidate rankings, voting intention, or election predictions. If the sample is too limited, no score is published.
Corrections and updates
Changes should remain visible.
Any published correction must explain what changed and why. Earlier versions remain in the editorial record. Requests can be submitted by email and are reviewed against the same evidence standard.
An index requires at least 20 qualifying items, 10 distinct authors or accounts, and three source types. It describes only the tone of the issue sample. It never measures candidate support or a likely election result.
How poll results are protected
A signed HttpOnly browser cookie helps limit each browser to one response, but it does not confirm a person’s identity. For an active poll, the server shows the breakdown only when this browser has a recorded response and at least 10 responses have been received. After a poll closes, final results become public only if that threshold was reached. The aggregate response count, field dates, status, and methodology remain visible without revealing choices. Clearing the cookie or changing browsers loses response recognition. To limit abuse on shared networks, no more than 20 new attempts per poll are accepted per hour for a poll-scoped hashed network identifier; no raw IP address is retained. If the limit is reached, no response is recorded and an approximate retry time is shown.
How poll context editorials are updated
A first context editorial may publish after a poll reaches at least 10 responses. We freeze the actual count at capture, then review new editions at 20, 30, 40 responses and so on. Material movement — such as a new leading choice or a shift of at least 10 percentage points at a threshold — may justify an update. These editorial checkpoints never make a poll representative or statistically significant.
How candidate responses are handled
Responses remain attributed to the candidate and separate from our editorial context. If a response is updated, the new version must clearly explain what changed.
How we review content
Facts, methods, polls, and editorial content are reviewed against their public sources and stated limitations.
