The Truth Filter

The reading room of the Institute had gained yet another addition: a sleek metal frame holding a series of glass slides, each etched with words like TRUE, FALSE, and UNCERTAIN. A small lever allowed one to “pass” statements through the device, and a printed slip emerged indicating the truth value.

Miss Elowen Stray approached it with a mix of curiosity and caution.

“And this measures… truth?” she asked.

Mr Blottisham stood beside the device, puffing up with pride.

“Exactly! Any statement goes in here, and the machine tells you whether it is true or false. Completely objective, of course.”

Elowen raised an eyebrow.

“Completely objective, you say? That’s quite a claim.”

Professor Quillibrace entered with his usual teacup, peering over the top of his spectacles.

“Ah,” he said, “the perennial dream: a machine that separates truth from falsehood without human intervention.”

Blottisham gestured proudly at the lever.

“You see, one simply feeds in a claim. The mechanism analyses it against a database of facts, logic, and evidence. Out comes the truth.”

Quillibrace tilted his head.

“And who decides what counts as a fact?”

Blottisham hesitated.

“Well… the database was compiled by experts.”

“And who chose the experts?” Quillibrace continued.

Blottisham frowned.

“I suppose… the institution.”

Elowen’s eyes lit up.

“So before a statement even reaches the Truth Filter, a great many human choices have already been made.”

Quillibrace nodded.

“Exactly. And yet the machine is often described as revealing truth itself.”

Blottisham crossed his arms.

“But it still outputs a verdict! The statement is true or false!”

Quillibrace sipped his tea.

“The machine outputs a result according to rules and references determined by humans. The verdict is contingent, not intrinsic.”

Elowen leaned closer to examine the printed slip.

“So truth, like the other properties we’ve discussed, emerges in relation to procedures, standards, and interpretation. The machine does not create it.”

Blottisham stared at the glass slides.

“So the filter doesn’t really measure truth itself?”

“No,” said Quillibrace gently. “It assesses a statement according to pre-established criteria. Truth is a relational achievement, not a standalone property of the sentence.”

Elowen nodded.

“And the debate about ‘truth in AI’ is often really about who designs the criteria, what sources are chosen, and which procedures are applied.”

Blottisham rubbed his chin thoughtfully.

“Well… I suppose I could add a dial for consensus confidence next.”

Quillibrace lifted his teacup with a faint smile.

“Indeed, my dear Blottisham. But remember: consensus, certainty, agreement—these are relations among observers, not properties magically encoded in glass and metal.”

The slides clicked faintly as the lever was returned to its resting position. For a moment, the room seemed less like a laboratory and more like a gallery of reflective judgments, where truth shimmered not in the filter, but in the interplay between claim, context, and construal.