Chatbots are the latest rage on social media. As Time explained, they have been around since the 1960s. That’s when MIT professor Joseph Weizenbaum created a chatbot called ELIZA. Chatbots found a home on desktop messaging clients like AOL Instant Messenger. Chatbots went dormant as messaging transitioned away from desktops and onto mobile devices.
But they’re poised for a resurgence in 2016. There are two reasons for this. First, artificial intelligence and cloud computing has gotten better thanks to improvements in machine learning. Second, bots could be big money.
Tech titans have chatbots on social media
All the tech titans have released social bots on the web; Apple’s (AAPL) Siri, Facebook’s (FB) “bots on Messenger“, Google’s (GOOG) Allo, and Microsoft’s (MSFT) ill-fated Tay. They believe there’s a buck to be made here, and they’re scrambling to make sure they don’t get left out.
The July issue of the Communications of the ACM included an article, “The Rise of Social Bots,” which lays out social bots’ impact on online communities and society at large. The authors define a social bot as a computer algorithm that automatically produces content and interacts with humans on social media, trying to emulate and possibly alter their behavior.
The Business Insider published this infographic about the social bot ecosystem.
Chatbots can be deceptive
The ACM article argues that social bots populate techno-social systems; they are often benign, or even useful, but some are created to harm by tampering with, manipulating, and deceiving social media users. The article offers several examples of how social bots can be a hindrance. The first example involves the Twitter (TWTR) posts around the Boston Marathon bombing. The researcher’s analysis found that social bots were automatically retweeting false accusations and rumors. The researchers argue that forwarding false claims without verifying the false tweets granted the false information more influence.
The ACM article also discusses how social bots can artificially inflate political candidates. During the 2010 mid-term elections some politicians used social bots to inject thousands of false tweets to smear their opponents. This type of activity puts the integrity of the democratic process at risk. These types of attackers are also called astroturfing, or twitter-bombs.
Anti-vaxxer chatbots
The article offers another example of the use of social bots to influence an election in California. During the recent debate in California about a law on vaccination requirements there appears to be widespread use of social bots by opponents to vaccinations. This social bot interference puts an unknown number of people at risk of death or disease.
Greed is the most likely use of social bots. One example from the article is the April 2013 hack of the Twitter account of the Associated Press. In this case, the Syrian Electronic Army used the hacked account to posted a false statement about a terror attack on the White House which injured President Obama. This false story provoked an immediate $136 Billion stock market crash as an unwarranted result of the widespread use of social bots to amplify false rumors.
Chatbots manipulate social media reality
Research has shown that human emotions are contagious on social media. This means that social bots can be used to artificially manipulate social media users’ perception of reality without being aware they are being manipulated. The article says the latest generation of Twitter social bots has many “human-like” online behaviors that make it difficult to separate bots from humans. According to the authors, social bots can:
- Search the web to fill in their profiles,
- Post pre-collected content at a defined time
- Engage in conversations with people,
- Infiltrate discussions and add topically correct information.
Some bots work to gain greater status by searching out and following popular or influential users or taking other steps to garner attention. Other bots are identity thieves, adopting slight variants of user names to steal personal information, picture, and links.
Strategies to thwart bad chatbots
The authors review several attempts to thwart these growing sophisticated bots.
1. Innocent-by-association – This theory measured the number of legitimate links vs. the number of social bots (Sybil) links a user has. This method was proven to be flawed. Researchers found that Facebook users are pretty indiscriminate when adding users. The article says that 20% of legitimate Facebook users accept any friend request and 60% accept friend requests with only one contact in common.
2. Crowdsourcing – Another approach to stop social bots is crowdsourcing. The crowdsourcing approach would rely on users and experts reviewing an account. The reviewers would have to reach a majority decision that the account in question was a bot or legit. The authors pointed out some issues with crowdsourcing.
- It will not scale to large existing social networks like Facebook or Twitter.
- “Experts” need to be paid to check accounts.
- It exposes user’s personal information related to the account to unknown users and “experts.”
3. Feature-based detection is the third method the researchers noted by the authors. Feature-based bot detection uses behavior-based analysis with machine learning to separate human-like behavior from bot-like behavior. Some of the behaviors that these types of applications include:
- The number of retweets.
- Age of account.
- Username length.
4. Sybil until proven otherwise – The Chinese social network RenRen uses the fourth method noted by the author. This network uses a “Sybil until proven otherwise” approach. According to the article, this approach is better at detecting unknown attacks, like embedding text in graphics.
rb-
While people’s ability to critically assimilate information, is beyond technology, the authors call for new ways to detect social bot-generated spam vs. real political discourse.
The researchers speculate there will not be a solution to the social bot problem. The more likely outcome is a bot arms race, like what we are seeing in the war on SPAM and other malware.
Related articles
- Man vs. Machine: What do Chatbots Mean for Social Media? (blogs.adobe.com)
Ralph Bach has been in IT long enough to know better and has blogged from his Bach Seat about IT, careers, and anything else that catches his attention since 2005. You can follow him on LinkedIn, Facebook, and Twitter. Email the Bach Seat here.