Laptop screen showing social media data_1400x300

Targeting Your Preferences

Research article

Micro-targeted campaigns (MTCs) are often viewed negatively as they draw on high levels of personal data. Now, researchers from LSE and the University of Oxford have set out to create a realistic voter model to understand the efficiency of MTCs, in the hope they can be applied to other areas of society, from public health messaging to encouraging pro-environmental behaviours.

Cognitive, political, and social psychological research are producing more appropriate models of relevant behaviours – as we gain better insight into the psychology of the citizens (belief revision, decision-making, etc.), campaigns can become more tailored and effective

Dr Jens Madsen

The use of data to inform and run political campaigning is an inescapable trend to help persuade and mobilise an ever-complex electorate. From the broad profiling of the population (society-level data) within a certain age range, to micro-targeted data (personal-level data) based on our likes, social media posts, and online habits that help target certain subsets of the voting population. Getting to know us is an extremely valuable exercise.  

The problem? Getting to know who we are and what makes us do what we do is messy, sometimes unethical. Previous studies have also tended to focus on specific case studies e.g. how targeted Facebook advertising was used in an election campaign, or the impact of YouTube over TV advertisements. These are valuable case studies, but given their unique contexts they fail to evaluate the fundamental impact of micro-targeted campaigning as a general tool that gives a more realistic picture of voters and their behaviours. 

This is why ongoing research from London School of Economics and Political Science (LSE) and the University of Oxford is so vital. To what extent are micro targeted campaigns effective in dealing with a diverse electorate?

Researchers Jens Koed Madsen (LSE) and Toby Pilditch (University of Oxford), used an existing Agent-Based Model (ABM) to explore the effectiveness of micro-targeted campaigns. In the model, the voters consider a series of political issues, ranging from climate change to the economy, and rank how much they care about each one. That is, voters not only differ in their beliefs in these policies, but also how much each policy influence their choice of political candidate. Further, the voters in the model each perceive competing political candidates as more or less credible when they were contacted by them. Finally, each voter is more or less likely to vote in a given election. This creates a very heterogeneous electorate that differ on policy opinion, policy weight, their perception of each candidate, and their likelihood to vote. Conversely, politicians in this model attempted to persuade the electorate that their stance on issues were the most appropriate.

What sets this model apart is that it includes two types of candidate: one who selects voters at random (that fits our society-level data model), and a candidate who uses a micro-targeted approach (that fits our personal-level data model) to select the most appropriate voter to contact on specific issues.

This approached enables the authors to test enables the fundamental efficiency of different outreach strategies for a given electorate. 

They found that data-driven, psychologically informed campaigns, MTCs, gain strategic advantage over more random, broad approaches, and show huge credibility in navigating an increasingly complex voter-space. And this applies also to the candidate too. If a candidate is perceived with little or no credibility, for example, but uses a micro-targeted campaign approach, they actually yield an advantage by being able to speak to a specific, and therefore gain advantage over, a subset of the population. 

With this model, the authors have been able to explore the benefits of MTCs against complex electorate issues and have taken a key step towards creating more realistic voter models. The authors note that while this research focusses on political micro-targeting, the principle of the method used could be applied to other domains, such as public health campaigns. 

Co-author, Dr Jens Koed Madsen says:

"We have to be careful not to extrapolate (one way or the other) and to sensationalise individual cases. If we want to legislate effectively and generate an equal platform for democratic debate, it is critical to understand how micro-targeting fundamentally works. For example, in the UK, campaign spending limits reduce the capacity for the kind of deep data mining we see in the USA where spending is unconstrained. As citizens in various democracies, we need to have a conversation on the access to and use of personal data when it comes to political campaigns. 

Micro-targeted campaigns are frequently portrayed negatively (for example, as acts of manipulation or potential breaches of personal information). However, these methods are fundamentally value-neutral. Of course, they can be morally questionable if they are used to effectively mislead or manipulate with people – however, they can also be beneficial if we learn how to better inform people of the benefits of the COVID-19 vaccine or encourage pro-environmental behaviour such as recycling. As with most insights, it is not the fundamental understanding, but the application that must be morally interrogated. 

Cognitive, political, and social psychological research are producing more appropriate models of relevant behaviours – as we gain better insight into the psychology of the citizens (belief revision, decision-making, etc.), campaigns can become more tailored and effective."

Behind the article

Pilditch, Toby and Madsen, Jens Koed (2021) 'Targeting “Your” Preferences: Modelling Micro-Targeting for an Increasingly Diverse Electorate' Journal of Artificial Societies and Social Simulation 24 (1) 5 <>. doi: 10.18564/jasss.4452

See also

Madsen, J. K. (2019) The Psychology of Micro-Targeted Election Campaigns, Palgrave Macmillan [i-xiv; pp. 395], ISBN: 978-3-030-22144-7, DOI: 10.1007/978-3-030-22145-4