Remove harmful terminology from report descriptions

[Mandy-cyber]

Context: Currently in the code the only thing we are filtering in the report data is the type of report. This should not be the case since typeOfReport will always have a finite number of options. Instead we need to be filtering reportDesc.

To-do:

• [ ] Create a list of harmful language that we do not want to accept in reports (e.g. racist, homophobic, etc terminology) .
  • Once this is done, comment on this same issue with the list you've come up with
• [ ] Create a filter function that looks for, and removes, those harmful terms from a given report

Considerations:

• [ ] Should a report not be allowed to go through if it has more than x% of harmful language?
• [ ] Are we allowing swear words? If so, are we allowing all or just some?

[psefia]

In terms of filtering, what are the ethical implications of filtering out a true story that had derogatory words / keeping a fake story that avoided using derogatory words? Maybe we should have a "flag" button for extra user safety (on top of the baseline derogatory words filtering) that allows users to flag reports that sound/are messed up. They can get sent to us, and we can evaluate it.

Also maybe instead of percentages, the filtering should just determine an x amount of times a certain derogatory word shows up in the report. We could have a tier list too, like

• Removed when seen once: racist/homophobic terminology
• Removed when seen twice: words like "hate", "suffer", "kill", etc
• Removed when seen 3 times: swear words

[Mandy-cyber]

Really great points, I agree! My only concern is with the tier list that might be a whole other bunch of code to implement. I definitely agree with that tier 1, we should for sure just remove as soon as we see it, but for the other two we can maybe just add a "Trigger Warning" in front of the report that says that the following report contains mentions of xyz. What are your thoughts on that?

[psefia]

That means there would technically be 2 lists: one for racist/homophobic terminology (once seen at least once, delete the report), and one for "threatening" (for lack of a better word), (once seen at least once, present a "Trigger Warning" with the mentions you talked about). The second one might be interesting to implement since it'd have to check for instances of every word in the second list (idk if runtime would be affected by that, I haven't taken algo yet lol)

[Mandy-cyber]

If the filtering and adding 'Trigger Warning' is done right as the report is made, the users and us really shouldn't experience any noticeable delay. Runtime would only be a serious concern (in my understanding) if either the report or our lists of terminologies, were long.