Hypothesis
Most folks re-share/retweet content because it's easy and it feels right to them. Providing a durable memory of times they shared things that were false/incorrect may help shape/revise ways they interact on social media.
Proposed intervention
Ask a user to follow an account, and/or upload their twitter archive. Comb through their tweets and identify things that are known to be false, and why. Present them with a single number score for where their shortcomings are. Chide/mock them privately, and suggest some steps to do better. Continue following their tweets and update their gullibility score. Tighten loop possibly?
Where Are the Lulz?
The lulz in this are entirely built around chiding yourself or others.
Concierge MVP
- Design a rubric which:
- maps individual tweets to truth/falsity
- provides a mechanism for aggregating truth/falsity values into a single score
- Get some folks to share their twitter archives.
- Try to run the rubric manually for some duration of time (or number of tweets)
Engineered MVP
- Build a bot that folks could follow/followback to collect their tweets.
- Build a thing that identifies topics of tweets
a. investigate mapping arbitrary tweets to topics which may have fact checks
b. build a smaller corpus of viral tweets by hand (since these are higher value targets, as they are shared by lots of people)
- extract the orientation/truth/falsity of the statement being shared
- Assign score to tweet
- aggregate scores
- collect petty insults to display to user, or things to lord over friends/family/colleagues
Associated research
{Links to papers and research supporting the hypothesis / intuition}
Hypothesis
Most folks re-share/retweet content because it's easy and it feels right to them. Providing a durable memory of times they shared things that were false/incorrect may help shape/revise ways they interact on social media.
Proposed intervention
Ask a user to follow an account, and/or upload their twitter archive. Comb through their tweets and identify things that are known to be false, and why. Present them with a single number score for where their shortcomings are. Chide/mock them privately, and suggest some steps to do better. Continue following their tweets and update their gullibility score. Tighten loop possibly?
Where Are the Lulz?
The lulz in this are entirely built around chiding yourself or others.
Concierge MVP
Engineered MVP
a. investigate mapping arbitrary tweets to topics which may have fact checks
b. build a smaller corpus of viral tweets by hand (since these are higher value targets, as they are shared by lots of people)
Associated research
{Links to papers and research supporting the hypothesis / intuition}