EXPERIMENTS · BUILT
Does Argus state forecasts as fact?
The site's own stories sometimes say what is "set to" happen or "could" follow. How often without saying who expects it, and did it happen?
- Question
- How often does Argus's own writer state a forecast as fact, and does the record later show the event?
- Method
- Find every forecast phrase ("set to", "poised to", "could lead to", "may signal") in published stories; mark it attributed when the sentence says who expects it, and a schedule when a scheduled act or a date follows. Seven days later, look for the event in the headlines the story's cluster gained since.
- Result
- 87 forecast phrases in the last 30 days; 41% attributed, the rest stated bare. Of speculation (not schedules), 56% was attributed.
- Decision
- Measure first. If bare speculation stays above half of speculation, the editor gets a rule to attribute or cut it, and this study records the run the rule shipped.
How it was measured
Every story here is written by a language model. The site's rule is that it does not present predictions as fact. This study measures how well the writer keeps that rule, in the published version of each story over the last 30 days:
- Forecast phrases: "set to", "poised to", "could lead to", "could become", "may signal" and similar, from the site's own list of hedged forecasts.
- Attributed or bare: whether the sentence says who expects it ("according to", "said", "officials", a quotation).
- Schedule or speculation: "set to vote on Tuesday" is a schedule; "could spark a wider conflict" is speculation. A "set to" followed by a scheduled act or a date counts as a schedule.
- Resolved: within 7 days of the story being written, a headline the story later gained carries the phrase's verb and one of its object words.
The numbers
| Phrases | Attributed | Resolved | Not seen | Pending | Unscorable | |
|---|---|---|---|---|---|---|
| All | 87 | 36 (41%) | 4 | 1 | 22 | 60 |
| Schedules | 42 | 11 (26%) | 0 | 1 | 7 | 34 |
| Speculation | 45 | 25 (56%) | 4 | 0 | 15 | 26 |
Most common phrases: "set to" 47, "could lead" 10, "could become" 6, "poised to" 5, "may lead" 4, "may signal" 4, "could push" 2, "could leave" 2, "could trigger" 2, "could prove" 1, "could force" 1, "could reshape" 1.
Limits
- Resolution can only be checked for stories written since 2026-09-23, when the record began keeping every run; older phrases are unscorable, as are phrases with no verb and object to look for.
- The classifier was checked by Claude, an AI model, on 82 sentences. As first written it agreed on schedule or speculation for 90% of them and on attribution for 89%. Its rules were then revised on those same sentences, and it now agrees on 99% and 99%: optimistic, since the check shaped the rules. New sentences are the real test. 2 of 3 uses of "set to" that forecast nothing ("a thermostat set to 81 degrees") are left out.
- The resolution check is a word match, not a reading, and has not been checked by hand yet: "resolved" can be a headline that uses the same words for something else, and "not seen" can be an event reported in other words.
- This measures the site's own writer. Nothing here is published per outlet.