I’m not seeing any mention of how they parsed out the data to account for reiteration of past reporting.
Let’s assume a different situation, one where Individual A slaps Individual B in a high school, and the incident is reported on in the schools newspaper. Initially, there is going to be a set of wording used: “B was the victim of an attack when A slapped them.”
You now have the first instance of “B was the victim” and “A slapped B”. Now let’s assume B slaps A the following day, resulting in a new newspaper story: “As we remember from yesterday, where B was the victim of a slapping by A. Today, B struck A.”
Now we have an instance of “B was the victim” and “A slapped B” but we have the first instance of “B struck A”. We’re two reports in, and using the methods described we have observed a bias.
Over the following days, similar retributive actions are carried out by both A and B. However, the root event doesn’t change. Each article is prefaced with something akin to “As we remember from previous days, where B was the victim of a slapping by A…” This doesn’t necessarily indicate bias, it can also be attributed to keeping the story grouped.
Is that what’s going on? I have no idea, and I’m not at all claiming that is the case. In fact, I don’t routinely pay attention to news from the BBC at all. But the lack of this being mentioned did make me wonder if it was factored in. That said, some of that language choice is obviously an issue. Especially “children” vs “people under 18”.
That said, if anyone sees something I missed in the github, I am more than happy to correct this comment to reflect my misunderstanding.