The DatAdat model of rethinking political participation in the Digital Era
Let’s get the elephant out of the room: Understanding voters requires data, ideally lots of it, and any effort to amass data is rightfully the object of suspicion. Given today’s political and legal climate, managing other people’s data properly is a massive challenge, and even well-meaning data managers who lack the requisite legal expertise may find themselves running afoul of the regulations that have become stricter in reaction to the Cambridge Analytica scandal.
Correspondingly, the first goal of any data management software worth its salt must be to ensure that, if handled properly, all the data are collected, stored and processed in a manner that strictly follows the relevant legal guidelines.
In engaging with our system, your voters should have the reasonable expectation that their data are managed in accordance with strict privacy regulations, while the clients can rest assured in the knowledge that they do not run the risk of inadvertently running afoul of data protection regulations.
Our software is designed to be fully compliant with what ranks among the strictest data regimes in the world, the EU’s General Data Protection Regulation (EU Regulation No. 2016/679). Essentially, if an EU client decides to leave the entire data collection and management process in our hands, then they need not worry about data law infringements. (For countries outside the European Union, we must work with the client to ensure compliance with the relevant legal requirements, although it must be pointed out that stricter standards than those imposed by the GDPR are rare, and correspondingly many clients may opt for voluntary compliance with the EU’s data regime.)
With respect to data retention, it will be assumed that any consent of data management is provided either for a fixed temporary period or until explicitly withdrawn by the user. Generally, it makes most sense to set the duration of a consent period as the most basic unit of “political time”, namely an election cycle. During this time, a registered visitor remains on the books even if their behavior is mostly or fully passive from there on out. Prolonged inactivity, however, will result in an implicit revocation of the right to use their data, while an express request by the user will result in an immediate deletion from the database.
What our model provides in the post-Cambridge Analytica environment is the best possible and legal exploitation of the restricted data space.
So we start with the assumption that you have visitors who have agreed to let you learn about their preferences: What’s your goal with them, and how do you get data about them in the first place? The goal is to use the data generated from your engagement with the voter described in the previous section and to translate these into an “understanding” of the voter, deeper insights that can be converted into actionable information.
Any voter will enter your system once they have made a clear indication of their willingness to share their data with you. As is standard today on most websites/apps which ask visitors to fill out or submit certain types of information, the voter in our case will be asked to consent to have their data stored, processed and analyzed by you or specific agents designated by you, including DatAdat. Whenever possible, the system will suggest that the visitor of our landing page or application uses a so-called social login to register, which is more convenient and makes it easier to authenticate the data. They will also be asked to share with us the data about their behavior on our Facebook pages, thus giving us an insight into their preferences as expressed in their activities on that specific social media. This is combined whenever available with the data collected through the landing page, where users are required to register and provide some basic information (primarily verifiable contact information), as well as the content-related information they share when we first engage them to reach out to us (e.g., by filling out a questionnaire, responding to consultation questions, signing a petition, or – most ideally – chatting with our chatbot).
Although we can complement these with data from your offline engagements with your voters (party membership lists, those who attend campaign or party events, etc.), the engagement-generated data provide the backbone of what we refer to as understanding the voters, our formation of insights about them.
Thanks to the underlying sophisticated algorithm that continuously analyzes the incoming information, the possibilities of understanding in this context are sheer endless. It allows us to analyze the larger group of voters that we have engaged in some way, any designated subset thereof, or even individual voters in the database.
Given sufficient data, the more encompassing overview of our voters will allow us to draw general conclusions about our “typical” voters in a variety of contexts, such as their cultural and media consumption patterns, their preferred hangouts, etc. Technically speaking, this dimension of understanding allows for the construction of so-called lookalike audiences, individuals that share many of the same traits as the given group and may therefore exhibit a higher than average propensity to be receptive to our messages.
Such information can help us in targeting our advertising or campaign activities in social media or even in physical locations.
At the same time, the database is not only useful for macro analysis but also in applying it to individuals to ensure that they can be reached with the most effective forms of communication (in other words, in discerning the particular channels of communication they prefer and are more likely to keep track of) and the kinds of contents that most interest them. If we know, for example, that a particular voter is almost exclusively interested in education, then it would be a mistake to inundate them with our policies concerning seniors. It would increase the risk of them losing interest, not necessarily because they disagree – matters of disagreement would arguably be relevant – but simply because they do not care and their very limited time to process information is taken up by what they consciously or at least subconsciously view as extraneous communication.
Crucially, the broader insights about the group overall also intersect with the individual-level data, for example in the form that the general insights about voter behavior also allow us to keep track of certain key statistics that indicate crucial information about the loyalty, interests, willingness to engage, etc. of individual voters. Most importantly, our aggregate information will reveal indications of decreasing activity/engagement. The algorithm will soon learn which specific indicators can be associated with attrition, sending warning signals if the engagement with the given individual is on the verge of becoming alarmingly anemic, providing the politician or party organization with the possibility of trying to revitalize the link to the voter in question.
What is vital for the system to work is ongoing engagement, as a source of new information, as a tool of fostering deeper ties and as a fundamental element of democratic discourse. One of the major benefits of the understanding generated by the system is that it will give us the tools to maintain engagement, to keep the relationship between politicians and their voters vibrant and pertinent. As long as that works, voters will also be readily mobilized because through their ongoing engagement they will see their own ideas and preferences reflected in the activities of the politician.