Open josiemundi opened 4 years ago
This seems like it could be an offshoot of Watson Tone Analyzer except for racial bias. I wonder if Watson TA could be trained with a data set specific to racial biases as a starting point for this idea.
I don't know much about Watson Tone Analyzer. I had a quick look. It seems like it is more oriented to getting customer feedback on products than tone of articles, but I agree it would be interesting if it could be a starting point for this.
It would already be possible to pick up a tone of sections of the article I imagine (like you say perhaps with a specific data set it could be trained for this specific use-case).
Theme:
Problem Statement 3: "Bias is learned and perpetuated in different ways (e.g. societal beliefs, misrepresentation, ignorance) that consequently create inequitable outcomes across all spheres of life."
Idea:
Check (draft) news content for statements that could be inherently biased or could cause bias in the public opinion. This could be used by people producing media content for news sites. Use some sort of NLP to scan content. If an article contains a statement that could be taken as fact or where racial bias may be present, the application would highlight it and suggest some additional resources to check and perhaps prompt to add in some statistical context to verify this has been done.
It could also be used by people reading the content (perhaps through a browser plug-in) in order to highlight areas of an article that may reinforce racism.
Uniqueness:
Something like this probably exists already (not sure) but potentially not for racial bias.
Impact if implemented:
Prompt creators of news content to be thorough in their analysis of whether the content they are producing is factually accurate and make them aware of the racial bias that is present in their writing.
Reduce reinforced racism from biased media content.