|Originally published in Inside the Influence Industry's Personal Data: Political Persuasion - How it works by Tactical Tech's Data & Politics team.|
Political tactics have long mirrored those of the marketing industry. In fact, virtually all of the methods explained in this guide were pioneered by for-profit companies before their arrival in politics. Though it is impossible to predict how exactly political campaigning will evolve in the future, the commercial sector and emerging areas of research and experimentation provide some hints as to what campaigns will be doing. This chapter explores several emerging technologies that use personal data and have gained some traction in recent political campaigns.
↘ Bots The use of political bots and computational propaganda in influencing online discourse, like trending hashtags on Twitter, has been well-documented and remains an active area of research. The rise of chatbots powered by personal data may lead to a more individualised version of this same phenomenon. As one researcher warns, 'in a few years, conversational bots might seek out susceptible users and approach them over private chat channels. They'll eloquently navigate conversations and analyze a user's data to deliver customized propaganda. Bots will point people toward extremist viewpoints and counter arguments in a conversational manner'.
Bots appeal to political campaigns for many reasons: they allow campaigns to respond to voter inquiries efficiently, help users navigate today's deluge of political information, avoid the risk of human error, work across platforms, and can provide personalised responses to the users chatting with them. They are also financially attractive: in April 2016, 'Facebook opened its hugely popular Messaging platform to bot developers', which is not only a cheaper option than bulk SMS but also enables campaigns to leverage a rich supply of Facebook data in the process.
Campaigns are also eager to use bots because of the personal data bots can gather. 'A chatbot can ask users to select certain options or to answer specific questions', one communications specialist explained, 'In this way, the bot creates an instant database and provides statistics about people's preferences, suggestions, and doubts.' In other cases, bots can target users with polls to measure responses to issues, allowing the campaign to collect further data.
Chatbots are also on the rise because they can be integrated into engagement efforts more seamlessly; they do not require any voter effort or initiative. After simply commenting on an article, a voter can be prompted to answer questions, sent to polls, asked to donate and more, all through bots.
Targeted political campaign bots are currently still quite basic. One startup in France, for example, built a bot that responded to any user message with one of 100 quotes from Donald Trump (and observed very high engagement rates). Another bot called 'Dein Selfie mit Van der Bellen' (Your Selfie with then-candidate and current Austrian President Alexander Van der Bellen) helped users add an image of Van der Bellen to their Facebook profile pictures.
The next generation of political chatbots are likely to be more sophisticated, especially as they glean more personal data. Campaigns are likely to make use of the same technologies and advances in natural language processing that Google, Amazon and Microsoft have used to make their bots more human-like. Adam Meldrum, an entrepreneur and specialist on the use of AI and chatbots in political campaigning, wants to use chatbots 'to create a more natural relationship with voters'. His vision, common in the industry, is to make chatbots 'respond like a human would'; that is, with enough improvement, bots can facilitate organic campaign interactions, unlike their current form - not much more than a 'glorified marketing site'.
In a similar vein, researchers at the MIT Media Lab are attempting to build AI bots that respond like their human counterparts. In the context of politics, using a bot 'knowledgeable about a candidate's positions, as well as their demeanor, the virtual conversation could allow voters to ask questions hyper-specific to their community, and receive targeted answers in return'. The idea behind the research is to probe whether our digital footprints reveal enough about our 'thoughts, interests, and personal identity' for an AI-powered bot to convincingly emulate us. If viable, campaigns would probably use this 'swappable identity' capability to 'influence opinion and generate excitement among people it views as likely supporters of their candidate'.
↘ Eye tracking Some political campaigns have also started refining their ads based on insights from eye-tracking research. In this work, a panel of people opt-in to having their eye movements recorded, either in a lab setting or using in-home devices. A blog post from Discida, a company that provides eye-tracking services to political campaigns, summarises the concept:
Eye tracking is an excellent technique that allows you to see exactly what is (and is not) noticed, and how much attention components of a piece are getting. If the voter doesn't look at the key images or words, then the piece fails to deliver its message. The benefit of eye tracking over other techniques, such as focus groups, is the response is automatic for the voter. They do not have to recall what they looked at, nor are they influenced to follow another member of the group or respond in ways that please the interviewer. Eye tracking provides that critical first few seconds of information -where do they look first and then where do they go next? Where do voters linger longest?
This technology allows political parties or candidates to tailor ads to voters for maximum impact. Researchers at the University of Vienna showed ads from the Austrian Green Party (liberal) and the Austrian Freedom Party (conservative) at the same time to liberal and conservative voters and observed their eye movements. The study found that people spent more time looking at the ads that aligned with their political views. While this may seem intuitive, a campaign could use it as rationale for many decisions to capture voters' attention, like crafting ads to voters' exact political leanings or publishing more polarising political ads. In turn, eye-tracking services will likely grow more personal as they optimise for increasingly granular groups. Eye-tracking technology contends, for example, that men and women look at different parts of ads.
↘ Cognitive computing Changes in technology have created ways to understand voters even more intimately, more quickly and supposedly more precisely. The marketing industry is exploring cognitive computing by conducting studies on the brain to understand how retention, emotion and attention are affected by the media we 'consciously and subconsciously' consume. A United States Postal Service blog post entitled 'What neuromarketing means for political campaigns' showcased research findings that 'ads that contain faces are remembered more than those that contain scenes or words'. The post also claims that physical campaigns (e.g., direct / physical mail) are better for raising candidate awareness because direct mail, when compared to other forms of media, demonstrated higher 'recall, desirability, and likability'. When combined with digital media, physical mail attracted 39% more attention than a single medium on its own.
↘ Internet of Things The sheer volume of information available to political campaigns is bound to increase dramatically. The number of connected devices worldwide is projected to surpass 30 billion by 2020 and campaigns are already positioning to extract as much value as possible from connected TVs, set-top boxes and media consumed online. In the near future, IoT smart speakers like Amazon's Echo, Google Home, robotic vacuums, smart beds and others promise to capture even more rich, behavioural data. Are members of a household concerned about safety issues? Campaigns no longer need to make predictions when an under-utilised smart alarm system can answer that question for them. One scholar predicted, 'for democracies, the Internet of Things (IoT) will transform how we as voters affect government---and how government touches (and tracks) our lives' via in-built sensors that 'never sleep'. Because many of these devices will live at home, the data we share with them will likely be even more intimate than the data we share with our devices today. Ultimately, these nascent developments are believed to advance campaigns' efforts to target voters as precisely and accurately as possible at scale.
In the United States: In November 2016, a Facebook Messenger bot created by @mssg was enlisted on behalf of three political groups; the Connecticut House Democratic Campaign Committee, the Pennsylvania Common Sense Political Action Committee and Bazta Arpaio, a community-driven campaign in Arizona, to vote Sheriff Joe Arpaio---accused of discriminatory practices---out of office. The bot asked voters for their address and returned the relevant voting location. While many websites to look up polling places existed at the time, @mssg's bot offered some advantages: a novelty factor, the convenience of a mobile-friendly experience and the opportunity to experiment.
The rollout of the @mssg bot was paired with a test comparing the bot to a simple website. As Beth Becker, who runs a company helping progressive causes connect with their supporters, reflected, 'collecting data in a conversation is a novel approach, and the initial results are amazing. We can collect any and all types of data from these users - address, email, phone number, date of birth, etc. And now that Facebook has enabled person to person payments via Messenger, it's just a matter of time before organizations can collect donations in this manner, too'.
In Canada: In the months leading up to the 2015 federal elections, performance marketing company Mediative assembled a small audience of five men and five women and reportedly used eye-tracking technology to 'tap into subconscious processes and decisions of an audience to understand which elements of the campaign websites' layouts trigger the fundamental brain circuits responsible to attention, cognition, and emotion'. The exercise used websites belonging to Canada's five major parties spanning the political spectrum and claimed to find, among other things, that men spent more time looking at logos, while women fixated more on family portraits. Some websites didn't help focus the eyes as much as others, suggesting areas for improvement. The blog post concludes, 'according to the findings of our study, it seems that Justin Trudeau is very popular among women; the conservative party's membership is more appealing towards men, the Bloc's layout can be confusing, the NPD is generating neutral emotions and the Green Party will attract more female voters. Joking aside, don't forget to vote and let democracy prevail!' While some may question the efficacy and validity of eye tracking, it reflects how campaign decisions are becoming ever more personalised.
While many of the technologies described here are not yet mainstream, they are likely to grow in popularity. If chatbots succeed in establishing rapport with voters, they will be difficult to distinguish from humans. Furthermore, there appears to be no way of knowing whether the ads or websites you consume have been enhanced using eye-tracking technology or cognitive computing tools intended to direct your eyes and mind to a specific message or image. The existence of connected IoT devices in your home does not necessarily mean that you are subject to precision targeting efforts by political campaigns, though the eagerness to leverage in-home devices for political purposes suggests this will change.
↘ If implemented in a privacy-respecting and transparent manner, an advanced chatbot could engage in a question-and-answer session specific to a voter's needs. For example, a voter could engage in a conversation about the impact of a potential policy on their business.
↘ Bots avoid the risk of human error in simple tasks like looking up polling locations for a given address.
↘ Ill-intentioned political players could use personal data to feed users personalised propaganda and promote extremist views. Bots without access to personal data have already been found to do so.
↘ Advanced chatbots could be made to mimic candidates' positions and promote spurious claims. Again in this case, voters may not know if they are communicating with bots or humans.
↘ IoT devices risk setting a precedent of dataveillance, the monitoring of online activities and digital actions for political insights. One academic commented, 'the IoT is essentially a massive surveillance network'.
↘ If techniques are combined, such as merging bots with micro-targeting, digital literacy challenges for voters and the complexity facing regulators will increase.
Author: Varoon Bashyakarla