verification-plugin
verification-plugin copied to clipboard
AFP-Medialab
Do we need an inner card for DBKF warnings?
When a page on the assistant comes back with DBKF hits, the user is presented with a warning box. This expands to another collapsed card which must also be expanded. Do we need the double wrapping or can we get rid of one layer?
Yes I agree, it's a bit unnecessary unless the previous fact-checks has a result which is shown as a second accordion. I implemented this when previous fact-checks needed to be moved from the now removed Credibility Signals card, but should be improved!
- example link with DBKF and previous fact-checks results: [find one]
- ... struggling to find a link which works. Setting threshold to 0.0 usually brings up results for testing: https://github.com/GateNLP/we-verify-app-assistant/blob/10f6a208380995e7ac0c29e3f5e9c55db0dd3e07/app/main/converters/kinit_converter.py#L12
Would need to also check the behaviour when an image is flagged with DBKF and it's contained in an accordion too.
- example link: https://pbs.twimg.com/media/F9CTQ2SXQAEYnPO.jpg
- example link with DBKF and previous fact-checks results: [find one]
-
actionnews3.com has previous fact checks (but no DBKF)
-
ancient-code.com has both (found by brute force search through the IFFY index
-
Here's another one just in case
-
Here's what it looks like with both of them open. I can see if there were several results for both that having both on one dropdown might be overwhelming.
Potential suggestions for improvement:
- Just have them all show up since most of the time we don't have super loads
- Have a few (e.g. 3 of each or up to six in total) come up by default and have a "show more/show less" option
- Make both options top level dropdowns in their own right since it seems to be quite rare to have both
What do we think?
For the URL Domain Analysis on the submitted link, we just have them all show up and it's looks fine. It's less for the user to click! The results (Warning, Mentions and Fact-checks) are separated by Chips so we could do the same: DBKF Text, DBKF Image and Previous Fact-Checks (or similar).
Let's go with that then. If the users are overwhelmed then I'm sure they'll let us know. Would it be helpful to change the title to something more meaningful that "Warning"? Perhaps "Database of known fakes" or similar?
Yes they will.
Let's go with that then. If the users are overwhelmed then I'm sure they'll let us know. Would it be helpful to change the title to something more meaningful that "Warning"? Perhaps "Database of known fakes" or similar?
Lets ask @Sallaa about a title change as that could be a big change for the users. And this leads to double checking the question again of whether the best place for the credibility signal Previous Fact-Checks is here.
Hi @rosannamilner @jmafoster1 thanks for pinging me. I agree that the less clicks, the better. Also "warning" by itself is probably not self-explanatory enough, which causes the user to have to click no matter what because they can't understand the warning.
A solution that you can consider is to use the mui alert with severity="warning" prop and have the results below in an accordion or a list, but not embedded in the alert. The warning could then have an explanatory message, and the accordion / list is always visible. This is more or less what we use in the Synthetic Image Detection (warnings) and Keyframes / KSE service (accordion for DBKF). So doing something similar will also help to have design consistency.
How do we feel about this?
I'm not sure. Could work better with the warning icon and title without the alert background colour?
This is a much better example of a link with both previous fact checks and DBKF hits. I've added it to the test links document too
Can we combine results for DBKF with Previous Fact Checks? Relates to comments in https://github.com/AFP-Medialab/verification-plugin/pull/917.
@jmafoster1 do you have a UI improvement in mind? As long as it is clear which is which, we could combine them.
I was really wondering whether it needs to be clear which is which since bother are essentially providing the same information? Do users of the assistant need to know which service (or both) have found the matches? If they do, please can you clarify the difference between what the two services provide? I know they work differently and are developed by different people, but they both seem to look through the same databases.
If we really do need to keep the two sets of results separate (or at least discernable), then we could have a pill saying which service(s) found each hit.