href="http://politybooks.com">politybooks.com
Acknowledgments
At this point, it’s hard to believe there was ever a time in my life when I hadn’t heard the word “bot.” The last decade of research and learning has been a thrilling journey, and I’m grateful for all the wonderful people I’ve met along the way. In no particular order, I want to express my deepest thanks to Tim Hwang, Marina Gorbis, John Kelly, Vladimir Barash, Camille François, Phil Howard, Clint Watts, Mark Louden, Frieda Ekotto, Helmut Puff, Roman Graf, Ralph Hailey, Rosemarie Hartner, Chou Changjen, Yao Yuwen, and Yauling and Joel for their support, guidance, and encouragement. I’m continually inspired by all the courageous journalists, activists, and researchers I’ve worked with over the years, especially my colleagues and friends in Taiwan – your work changes lives. A huge thanks to my friends Sam C., Nate, Amanda, Ike, Anj, Sylvia, Quin, Renata, Samantha, Jake, Jackie, Skyler, Trevor, Jane, and Doug for making life so full and always being up for an interminable conversation. My co-author, Sam Woolley, you’ve been an incredible friend and colleague, and I’m already looking forward to our next project. Lastly and above all, I’m most grateful to my wonderful family – Mom, Dad, Grammy, Mark, Ben, Britnea, Franki, Rocco, Benni, Murphy, Andi, and all the Monacos and Carmacks. Your love has made me who I am. There’s no better family on Earth.
Nick
First and foremost, I would like to thank my family for their constant, enthusiastic, support of my work. Without their encouragement, advice, and love I would never be able to do what I do. To Samantha, Pip, Mum, Dad, Oliver, Justin, Daniela, Manuela, Basket, Mathilda, Banjo, Charlie, and the Woolley, Donaldson, Loor, Shorey, Westlund, and Joens families – a sincere thank you for everything. To all of my friends – particularly Nick Monaco – thank you so very much for all of the learning and laughs. You make each day fun and inspiring. I’d also like to thank the members of my research team, the Propaganda Research Lab, collaborators at the Center for Media Engagement, and colleagues at the School of Journalism and Media and Moody School of Communication – all at the University of Texas at Austin. Finally, I’d like to thank the organizations that support my ongoing research, particularly Omidyar Network, the Open Society Foundations, and the Knight Foundation.
Sam
Abbreviations
AI – Artificial Intelligence
ANT – Actor Network Theory
ASA – Automated Social Actors
CUI – Conversational User Interface
GPU – Graphics Processing Unit
GUI – Graphical User Interface
HCI – Human–Computer Interaction
HMC – Human–Machine Communication
IO – Information Operations
IRC – Internet Relay Chat
ML – Machine Learning
MT – Machine Translation
MUD – Multi-user domain, multi-user dungeon
NLP – Natural Language Processing
RES – Robot Exclusion Standard
STS – Science, Technology, and Society studies
1 What is a Bot?
The 2020 United States presidential election was one of the most impassioned in the country’s history. President Donald Trump and his Democratic opponent Joe Biden both contended they were fighting for nothing less than the future of American democracy itself. The election brought with it several events rarely seen in the history of American democracy – an election held in the middle of a global pandemic, citizens’ storming of the US Capitol, and attempts by a sitting president to overturn the results of a free and fair election. Unprecedented events weren’t only taking place offline, however – social bots, or computer programs posing as humans on social media sites such as Twitter and Facebook, were beginning to use artificial intelligence (AI) techniques to fly under the radar of security teams at social media platforms and target voters with political messages. One of the leading bot detection experts in the US bluntly admitted,
Back in 2016, bots used simple strategies that were easy to detect. But today, there are artificial intelligence tools that produce human-like language. We are not able to detect bots that use AI, because we can’t distinguish them from human accounts. (Guglielmi, 2020)
But bots were not only carrying out covert, deceptive, activity online in 2020. Working with amplify.ai, the Biden campaign deployed a chatbot to interact with users on Facebook messenger and encourage users to vote. This bot’s intent was not to deceive – it would reveal that it was not human if asked – rather it was a means of using AI techniques to try to boost the get-out-the-vote efforts. Amplify.ai’s bots helped Biden reach over 240,000 voters in fourteen states in the three weeks leading up to election day (Dhapola, 2021; Disawar & Chang, 2021). Bots’ activities in the 2020 election illustrated the dual nature of the technology – whether bots are “bad” or “good” for society depends on how they are designed and used.
Until recently, the word “bot” was fairly obscure, used mostly in arcane discussions in the academy between scholars, and in Silicon Valley meeting rooms full of computer programmers. The year 2020 was, of course, not the first time bots had been deployed to participate hyperactively in online political discussion in the US. The November 2016 presidential election was the one that gave bots a household name, both in the US and around the world. Journalists and researchers documented the underhanded automated tactics that were being used during that contest to promote both candidates. For many, this was the first time that they realized that political discussions online might not have an actual person on the other end – it might be a piece of software feeding us canned lines from a spreadsheet on the other side of the globe. Now, we can’t seem to get that idea out of our heads. These days, social media users quickly label any antagonistic arguer on social media a “bot,” whether it’s a troll, a disinformation agent, or a true bot (an automated account).
But before bots became a notorious byword for social media manipulation in 2016, they were already a central infrastructural part of computer architecture and the internet. Many bots are benign, designed to do the monotonous work that humans do not enjoy and do not do quickly. They carry out routine maintenance tasks. They are the backbone of search engines like Google, Bing, and Yandex. They help maintain services, gather and organize vast amounts of online information, perform analytics, send reminders. They regulate chatrooms and keep them running when users are fast asleep. They power the voice-based interfaces emerging in AI assistants such as Apple’s Siri, Amazon’s Alexa, or Microsoft’s Cortana. They carry out basic customer service as stand-ins for humans online or on the phone. On the stock market, they make split-second decisions about buying and trading financial securities; they now manage over 60 percent of all investment funds (Kolakowski, 2019). In video games, they run the interactive agents known as non-player characters (NPCs) that converse with human players and advance storylines.
Other bots are malicious. They amplify disinformation and sow discord on social media, lure the lonely onto dating sites, scam unsuspecting victims, and facilitate denial-of-service cyberattacks, crashing websites by overloading them
