AI's upside/downside

The increasing use of artificial intelligence (AI) is both exciting and troubling. There is tension between the possibilities of convenience and efficiency (augmenting human activities) and ethical concerns, such as privacy and consent. The trick, it seems, is balancing the enthusiasm for new technologies with the right amount of deliberation about who benefits most — along with consideration of longer-term consequences, such as how these technologies might be misused.

Some events last week illuminated the complex upside/downside of AI’s progress:

A still of the demonstration video for Google Duplex, a new virtual assistant. Click the image to view the demo.

Google Duplex: Though Google’s new Android P mobile operating system is full of changes driven by AI — including a new Google News algorithm that makes a number of judgments (including about the credibility of news reports and about a user’s likely level of interest in them) — the announcement and demonstration of a virtual assistant called Google Duplex making a hair appointment stole the show. It also sparked jokes — and significant concerns.
 

• Discuss: Is it unethical to have an automated assistant call a business and interact with a human on your behalf? Is it unethical for such assistants to use fillers (such as “um” and “uh”) that make them sound more lifelike? Is it important for lifelike bots to disclose the fact that they are machines?
• Idea: Have students show the Google Duplex demo video to friends and family, then document their reactions (by using a questionnaire, for example), then collect and create a visual to show the results.
• Related: “No one knows how Google Duplex will work with eavesdropping laws” (Sarah Jeong, The Verge)

Facebook news feed tweaks: In January, Facebook announced that it would be changing its news feed algorithm to downgrade “public content” from businesses and brands and upgrade content from friends and groups. The move caused significant concern that quality news sources might be downgraded and experience large drops in traffic.

A recent analysis suggests that whatever changes Facebook has made since January don’t appear to be hurting mainstream news outlets, but do appear to be dropping the traffic and interaction of highly partisan, less reliable publishers, like Gateway Pundit.

• Note: The changes to Facebook’s news feed have also sparked accusations of partisan bias from both conservatives and progressives.
• Related: Facebook recently added — and quickly removed — a new “Page History” feature that allowed users to see more information about organization pages on the platform, such as when they were created, how many administrators they have and where those people are based. Facebook says it was an early (and temporary) test for users in Canada and Ireland. Reporters at BuzzFeed News, which has a significant presence in Canada, managed to capture information about a number of partisan pages before the feature disappeared.

Predictive suggestions still problematic: Jonathan Albright, the research director at Columbia University’s Tow Center for Digital Journalism, has again captured and shared objectionable search suggestions — this time on YouTube:
 

• Note: This topic was covered in more detail in the March 5 issue of The Sift.
• Act: Challenge your students to join Albright — and others in the above thread — in documenting questionable predictive search suggestions and sharing them publicly.

Russians target race

Last Thursday, the Democratic members of the House Permanent Select Committee on Intelligence released over 3,500 ads that were placed on Facebook and Instagram by Russian operatives between June 2015 and August 2017. More than half of the ads made explicit references to race in ways that sought to deepen racial tensions in the United States. A third of the ads, including some that did not directly refer to race, invoked the issues of crime or policing (25 percent) and immigration (8 percent).

Some ads carried messages designed to suppress voting in communities of color, while others sought to exploit and foster a distrust of news media.

Two of the more than 3,500 posts that were created and placed as paid ads on Facebook and Instagram by Russian operatives between June 2015 and August 2017.

The ad data released also provide insights into how Russia used Facebook’s Ad Manager audience tools, along with mining the Facebook Page followers of other ideological pages, to target specific groups of users with messages. For example, it created a fake Facebook Page called “Back the Badge” and then paid 110,058 rubles (about $1,785) to promote it to Facebook users who had showed support for police.

An ad promoting a Russian propaganda Facebook Page called “Back the Badge”
was the top performer of the 3,517 ads that were released. It reached 1.3 million users.

• Note: Only 10 percent of the ads contained messages that were election-related, and only about 100 stated support for either Donald Trump or Hillary Clinton.
• Also note: This series of graphics from USA Today might be useful in the classroom.
• Idea: Ask groups of students to download one portion of the ads released by the Intelligence Committee Democrats (they are organized by year and quarter) and conduct an analysis. Then discuss as a class:
  • Discuss: What specific themes emerged in the posts you reviewed? What social issues were common? What did you think of the messages themselves? Were there any to which you had a strong reaction? Do you think you were exposed to any of these on Facebook or Instagram?