The Expert on Your Deadline
It’s 4 PM. A reporter is wrapping up a story on a proposed soda tax in schools and needs a scientist to explain how sugary drinks affect children’s health. Three names emerge, all with impressive publications, university titles, and prompt replies. But who has time to check 40 papers for beverage industry funding? So the quote goes in, and the expert is presented as neutral. That’s the gap TieTrace is built to fill.
A story that already happened
In 2015, reporters revealed that the Global Energy Balance Network, a group of scientists who argued publicly that exercise mattered more than diet in the obesity debate, had received money from Coca-Cola ( CNBC, republishing the New York Times ). Several of those scientists had been quoted as neutral experts for years.
Not every dollar in that story was hidden. Part of the trail sat in plain sight, in funding statements and affiliation lines of published papers. The problem was not secrecy. The problem was that nobody on deadline can read all of it.
What TieTrace does
A journalist gives TieTrace three things: the company in the story, the topic, and the experts they plan to quote. TieTrace searches a research graph built from OpenAlex , an open catalogue of scholarly papers, and comes back with evidence for each expert.
- Tier A1: The expert wrote a small paper (25 authors or fewer) that the company funded, or listed a company affiliation on one.
- Tier A2: The tie runs through the company's foundation, a bottler, or only a very large multi-author paper.
- Tier B and C: The expert works closely with people who have direct ties. TieTrace shows this as background and labels it clearly as unvalidated.
- No tie found: Nothing turned up within three co-author steps.
Every Tier A claim comes with the paper's DOI, its year, the funder, and whether the paper is on the story's topic. A reporter can click the link and read the funding statement in under a minute.
The words we chose not to use
TieTrace never says "independent". It never says "conflicted". Those are verdicts, and a database cannot deliver a verdict.
Instead, it says "ties found in the research record" or "no tie found", and the second phrase always comes with its limits attached. A typical answer reads like this:
No tie found within 3 co-author hops, papers from 2021 to 2026. 56% of topic works in this graph list a funder (this expert: 92% of 26 works).
That last number matters. If only half the papers in a field record who paid for them, then "no tie found" might simply mean "no funding recorded". TieTrace tells the reporter how much the silence is worth, expert by expert.
It remembers what the newsroom already checked
Newsrooms check the same experts again and again. TieTrace keeps a private memory for each newsroom. When a colleague opens a new story next week with the same scientist, TieTrace says something like "already checked on Oct 16, tier A2 via Coca-Cola Foundation funding", and shows what has changed since.
One newsroom can never see another's memory. We tested this by trying every tool with another newsroom's identifiers. Every attempt came back empty.
A human stays in charge
Two actions in TieTrace can change the record, and both need a person to approve them:
- Ruling out a paper. If a reporter decides a paper is out of scope, TieTrace re-runs every saved check in the newsroom and shows exactly which results changed.
- Approving a sourcing note. An AI drafts a short note listing the evidence, a separate checker verifies every claim against the database, and then a human approves it before it is exported with live DOI links.
The AI can propose. It cannot publish.
Experts are also shown as "Expert #1", "Expert #2" by default. Names appear only when a case is explicitly marked public, so a draft investigation cannot leak who is being looked at.
Does it work?
We tested TieTrace on the case that inspired it. Of the four publicly confirmed leaders of the Global Energy Balance Network, TieTrace placed all four in Tier A, each with the DOIs to prove it. A fifth name we could not confirm from a public source, so we did not guess.
Then we built a second case from scratch, Philip Morris and lung cancer research, to make sure the tool was not hard-wired to one company. It worked end to end. It also reported, without being asked, that only 37.5% of that graph's topic papers list a funder at all. A tool that tells you when its own answer is weak is a tool you can trust with the strong ones.
What it cannot see
We want to be plain about this:
- Gifts to universities are invisible. Much of the 2015 money went that way, and paper funding fields do not record it.
- Funding data is thin. Many papers never list a funder.
- The graph is a sample of the most-cited papers in each year, not the whole field.
TieTrace does not replace reporting. It replaces the hour of scrolling that comes before it, and hands the reporter the right question for the phone call.
Why this matters
Trust in expert voices depends on readers knowing who those experts are. A reporter who quotes a funded scientist without saying so is not lying, but the reader is still misled. The evidence to prevent that usually exists. It is just scattered across thousands of PDFs.
TieTrace is open source under the MIT licence: github.com/Aaryan-Degama/TieTrace . Research data from OpenAlex (CC0).
By Abdul Azeem. Built for FalkorDB Graph Hacks, 2026.
Enjoyed reading this content? Let the author know!
Your likes, comments, shares, and saves help creators reach more builders.
Loading recommendations
Loading article