For your consideration: FactCheckNI verdict ratings

What does it mean for something to be “true”, “correct”, or “accurate”? Can anything be absolutely true or completely false? Is there any difference between being correct and being accurate?

Most fact-checking projects employ a verdict rating system. The initial policy at FactCheckNI was to avoid using an explicit rating system. Instead, we suggested that the reader draw his or her own conclusions from the concise summary provided in each article that fact-checked a claim.

We based this policy on our motivation to inform and educate, rather than to determine a judgement, or intention, on the “truthfulness” of the claim.

Researchers recently asked us to categorise the results of our claim investigations. We paused to assess how we had utilised a range of words prominently in our fact-check conclusions and discussed whether we could tighten up our usage to increase comprehension and reduce confusion.

Many of our article conclusions contained the term “accurate” or “inaccurate”. What do those terms mean?

In data sciences, there is an established procedure to determine a margin or error (e.g. +/- x%); if something falls outside this margin, then you can question its accuracy. If there are five people seated in a room, then – in terms of accuracy of numeric estimation – the answers “4”, “5”, or “6” are all accurate. There is no need to determine a status of “less than accurate” or “more than inaccurate”.

In the case of qualitative claims, how close is the claim to the evidence? For example, you can check the audio/video recording or official transcript against what was reported to be said. Perhaps it corroborates (“accurate”). Perhaps it almost corroborates; in such cases, we will say “accurate with considerations”, pointing out the inaccuracies but not deeming the whole matter inaccurate. This can work the other way round, too — “inaccurate with considerations” — when there’s an element of the claim that we wish to underline.

“With considerations” is still a qualifier, such as “mostly”/”almost”/”partly”/”half”. But it is not a statement on the degree of “true”/”truth”/”correct”/”right”-ness of the claim. Just that there are some aspects that we wish to draw to your attention for consideration.

And then there are those claims where we don’t come to any conclusion on accuracy or inaccuracy. This may be because of insufficient information to substantiate or rebuke the claim. Thus, we will apply a conclusion of “unsubstantiated”.

The following verdict ratings will be applied to FactCheckNI fact-check articles:

  1. accurate (evidence corroborates; within margin or error)
  2. accurate with considerations (qualifications)
  3. unsubstantiated (insufficient information)
  4. inaccurate with considerations (qualifications)
  5. inaccurate (evidence does not corroborate; outside margin of error)

This effort is to try to establish a system for our claim verdicts that is as objective as possible.

We still argue that it is up to the reader to determine how much weight to apply to any of our article conclusions. Is a misapplied piece of satire better or worse than a state agency misreporting economic data? We have our views. It’s up to you to defend yours.

Our hope is that other fact-checking projects see the benefit of a more objective-based system of claim verdicts — to take some subjectivity out of scrutinising facts and to spend more time on other crucial aspects of our work, such as developing a greater appreciation of media literacy and critical thinking.