Gover, D., & Sharevski, F. (2021). Two Truths and a Lie: Exploring Soft Moderation of COVID-19 Misinformation with Amazon Alexa. arXiv preprint arXiv:2104.04077. https://arxiv.org/abs/2104.04077?context=cs.CR

University of Michigan - Ann Arbor

- Researchers studied how Twitter users’ perceptions would change as a result of spoken soft moderation through Amazon Alexa because people generally do not question Alexa’s knowledge
- Alexa has this function where it can distort information and tell you something that is completely different than the intended truth as a way to mislead users. The original intent of this skill was to provide two different perspectives on government regulations.
- The study seeks to understand whether Alexa’s skills can also be used to reduce and diverge user attention from misinformation through the execution of 3 situations – Alexa informs users of the warning label and repeats the misinformation tweet after; Alexa first repeats the misinformation tweet then reads the warning label; Alexa does not say anything about the warning label and simply reads the tweet containing misinformation to the user 

Two Truths and a Lie: Exploring Soft Moderation of COVID-19 Misinformation with Amazon Alexa

- After users ask Alexa to access the Twitter Reader and Alexa receives a confirmation from the user to read their Twitter stream, Alexa gets all the tweets with warning labels and starts reading them, only focusing on plain text and ignoring symbols and characters such as the hashtag
- In the first case, Alexa informs users of the warning label and asks them if they would like Alexa to keep reading the tweet. If a user answers no, Alexa will ask to read the next tweet in the stream, otherwise Alexa ends the conversation and exchange. The study on the accuracy of tweets compared user perceptions regarding tweets with vs. without warnings and Alexa mentioning the warning before vs. after reading the tweet contents
- In the second case, Alexa simply reads the warning label after reading the whole tweet. The study on hesitancy and political leanings focused on whether a user’s political affiliations will impact their perceptions of the accuracy of the tweet contents through Alexa’s verbal warnings and reading
- Participants recruited from Amazon Mechanical Turk and had to be at least 18 years old, uses Twitter and Alexa, and has seen at least one tweet related to COVID-19 vaccines
- Experiment consists of participants answering questions – one about perceived accuracy where each person rated the content on a Likert scale (not accurate at all, slightly inaccurate, slightly accurate, or very accurate), a second on their intent to get the vaccine (yes, no, or I don’t’ know), and a final one about their political affiliation (Democrat, Republican, or Independent)
- Participants were randomly assigned into one of 6 groups – repeated tweet with warning; repeated tweet without warning; warning before repeated tweet; repeated verified tweet; repeated verified tweet with warning; warning before repeated verified tweet

Two Truths and a Lie: Exploring Soft Moderation of COVID-19 Misinformation with Amazon Alexa - Methodology

- The experiment consisted of 304 participants and ran from January to February of 2021
- Regarding the accuracy of tweets, most users simply ignored the warnings after Alexa repeated tweet contents and rated tweets with warnings tags as “somewhat accurate”
- Warnings before Alexa repeated tweet contents were more effective because these users generally rated the tweets as “not very accurate,” suggesting that warning covers are more influential than warning tags in convincing users of misinformation 
- People who would like to get the vaccine rated misinformative tweets as “inaccurate” whereas people who do not necessarily want to get the vaccine perceive the misinformative tweets as “somewhat accurate” 
- Democrats and Republicans regarded tweets with warning covers and tags as “not so accurate” while Independents rated them as “somewhat accurate”

Two Truths and a Lie: Exploring Soft Moderation of COVID-19 Misinformation with Amazon Alexa - Results

- Overall, when participants heard warning covers spoken by Alexa, they seemed to pick up on that because they saw Alexa as an impartial third-party who was simply conveying the truth, and hence this effort was effective in informing users of misinformation
- As for political affiliations, Independents had a greater chance of being skeptical about the warnings repeated by Alexa, whereas Democrats and Republicans generally believed what they heard from Alexa as long as it matched their beliefs 
- When Alexa informs participants of a warning tag before repeated the tweet content, participants were more likely to believe in the misinformation, regardless of whether the tweet content contains misinformation or not 

Learn Before

Related