You are here
Voter profiling in the 2017 Kenyan election
Image source: AFP
Earlier this month, the Kenyan daily The Star reported that UK-based data analytics firm Cambridge Analytica had been quietly contracted by President Uhuru Kenyatta’s party in a bid to win himself a second term in office. State House officials were quick to deny the claims, while the company itself issued no comment.
Cambridge Analytica has exploded onto the scene following revelations that its psychometric profiling techniques were used and reportedly played a role in both the US election and the UK Brexit referendum. The company has been in the limelight ever since, and the UK information commissioner is launching an investigation into the way political parties target voters through social media.
The news that Cambridge Analytica has reportedly been hired by an increasingly unpopular president (according to some polls) to win a very close, tense election, does not necessarily give case for alarm.
This is not the first time that Kenyatta’s administration has carried out opinion polling to hedge its chances. In fact, Cambridge Analytica worked on behalf of one of the candidates in the 2013 election, claiming to have implemented the largest political research project in Africa. The Kenyan government has strategically called on the services of UK firms in the past. London-based public relations Agency BTP Advisors, which is reportedly working with Cambridge Analytica over the next few months, represents the Kenyan government, including in interactions with the UK Parliament. BTP Advisors, also claims credit for building scepticism about the International Criminal Court in Kenya, during the ultimately fruitless attempt to prosecute President Kenyatta for alleged involvement in crimes against humanity following the 2007 elections.
But in an election context, profiling and microtargeting of voters is particularly problematic. Kenya’s presidential election this year is no ordinary one. That the Kenyan president’s Jubilee party would commit a reported 6 million USD to Cambridge Analytica for a mere 3-month project contract so soon to the polling date, and with such secrecy, gives cause for further scrutiny.
A short (recent) history of a bad idea
Political campaigns frequently rely on data, from constant polling, to finding out where to hold rallies or how to communicate to undecided voters.
As people’s lives become increasingly digital, the volume of data available to political campaigns and marketers increases, and so does the granularity of data available. Everything we do is becoming traceable: the websites we visit, who we talk to (and how often), where we have been, and what we care about. All of these data can be used to make sensitive inferences and predictions about people. Personality traits, for instance, can be predicted from Facebook likes, or even phone metadata. Behavioural data can also reveal someone's sexual orientation - even though an individual has never decided to consciously reveal this information anywhere online.
Access to such data means that political campaigns and marketers are able to create detailed profiles of individual voters and entire populations. As early as 2008, for instance, the Obama campaign had reportedly gathered as many as one thousand variables about each voter drawn from voter registration records, consumer data warehouses, and past campaign contacts to better target its messaging.
Such profiles include sensitive information, such as how likely an individual is to vote at all, their income bracket, interests, health, opinions, habits, or even predicted future behaviour. It also means that messages can be microtargeted to individual voters, depending on how likely they are to be persuaded.
Enter Cambridge Analytica. The company markets itself as unique because it does not simply predict individuals’ interests or future behaviour, it also creates psychometric profiles (in addition to the plethora of other information that can be obtained and predicted from the data traces we leave behind). Psychometric data include personality, that is information about how an individual scores on a number of measures, from neuroticism, to openness or extraversion.
Data is not just important to target individual voters. Online communication has radically transformed the way that people learn about the news, including political campaigns and issues. Some suggest that this has enabled new forms of automated propaganda to spread in ways we do not fully understand - from fake news to political bots on social media. While Cambridge Analytica specialises on political campaigns, it was created by SCL Elections - a company that focusses on data-driven communications for governments and military organizations worldwide.
What we do know from the Snowden revelations is that Western intelligence agencies have considerable interest in harnessing these new possibilities. The British signals intelligence agency GCHQ employs a unit called JTRIG that uses insights from social science, such as psychology, to engage in online manipulation and (mis)information campaigns.
More red flags than a bullfighting convention
Collecting data about millions of Kenyans to profile, target, and persuade voters on behalf of the Kenyan president is a strikingly bad idea.
Let us count the ways.
1. In Kenya, ethnicity has become political. Very political.
The amplification and politicization of distinctions between Kenya’s over 70 ethnic groups results from a long history of manipulation by elites for personal and group gain – a history in which Kenya’s colonizer, Britain, played no small part. (See here and here, for example). Post-independence Kenya broadly saw the emergence of a political elite of Kikuyu ethnicity, for a number of complex reasons including favourable access to land in Kenya's fertile rift valley, while ethnic Kalenjin and Luo political elites also consolidated their own political support among co-ethnics in an attempt to access state power.
Fast forward to 2007. Widespread dissatisfaction over the re-election of president Mwai Kibaki in December against contender Raila Odinga erupted as extreme violence expressed often in ‘tribal’ language and targeting. At least 1,100 people were killed and thousands forced to flee their homes; many of these were victims of the now-banned Mungiki citizens' militia, which espouses a virulent form of Kikuyu nationalism. The International Commission of Inquiry on Post-Election Violence report attributes much of the violence factors including many voters’ perception that it was essential that the ethnic group from which they come to win the Presidency in order to ensure access to state resources.
“Tribalism” is thus both the thing that should not be discussed – the National Cohesion and Integration Commission assiduously monitors ethnic hate speech – and also a resilient conceptual framework through which economic, social, and political grievances, particularly against the President's Kikuyu ethnic group, are expressed
Profiling individual Kenyans – of the kind that Cambridge Analytica’s technology would do – is particularly sensitive in this context. In Kenya, someone's name is often all you need to discern their ethnicity. Gathering such personal data on millions of Kenyan citizens is highly problematic, especially since it is unclear how such data will be stored and who will have access to it. Such questions become even more pressing if the election result is contested. The Cambridge Analytica affair has already fed into a popular perception that the ruling party has allegedly mobilized state resources in an attempt "rig" the election, as expressed by many commentators.
In short, the issue is already having an impact, and it’s not pretty: “[Cambridge] Analytica is feared for extreme scaremongering and fearmongering…” opines one opposition group on Facebook, for example. “Jubilee is intending to make this election the dirtiest, most hate-filled campaign to retain power.”
2. Corruption and lack of transparency on political spending.
Kenya suffers extraordinary levels of corruption, particularly in the public sector. Transparency International ranks Kenya among the top 30 most corrupt countries globally, according to its Corruption Perceptions Index. Scandal after financial scandal has rocked the Kenyatta administration, and the Mwai and Moi administrations before it.
The opposition has long accused the incumbent administration of taking extraordinary advantage of access to state funds to cement its power. For its part, opposition candidate Raila Odinga, too, has vowed to spare no expense in setting up an alternative vote counting system, to ‘rig proof’ the results. The secrecy and denials around the reported Cambridge Analytica deal, which at its most benign is another example of a candidate mobilizing data to their cause, do little to dissuade these tensions. Within the context of pervasive state corruption in Kenya, the project risks adding to the gaping absence of faith many Kenyans feel in the democratic process.
3. Kenya has no data protection law.
Let’s imagine that Cambridge Analytica do for the Jubilee party what they say they do for other campaigns. If CA's work is just slightly similar to what we have learned about their work in the US and Europe, they will be creating a database of highly sensitive information about a considerable share of a country's population. We don't know what data sources they will use, or whether they will be doing psychometric predictions. In this context, it is important to note that personality has been predicted from phone metadata - not just Facebook likes. The possibilities are endless. Could such a database contain information about ethnicity, political leaning, and who knows, sexual orientation or even health data? All of this happens in a politically volatile country, as part of a tight and tense election. Will Cambridge Analytica share or sell this data? Will they store it securely? Could the data be leaked, hacked, or stolen? Could it be accessed by the government? By the President?
There is no comprehensive guidance on how data collected and processed in Kenya should be stored, retained, and protected – and this is worrying.
The Kenyan government has expanded its powers of surveillance and data collection. It passed Security Laws (Amendment) Act. It is steaming ahead with the Cybersecurity Bill. It has put in place a number of serious cybersecurity initiatives (more on that from us later). But when it comes to data protection, those initiatives have lost steam – the Data Protection Bill has been languishing since 2013.
4. Rampant unaccountable spying.
The intelligence services have a proven record of exceeding their legal mandate to spy on Kenyans, according to our latest investigation on surveillance in the country. Again, this does not inspire confidence in the government’s desire to “play by the rules” of democratic processes, such as elections.
5. Finally, the election is slated to be a very close and possibly contested.
Tensions are high. Following April's primary elections, there have already been clashes between supporters of rival parties, which President Kenyatta recently condemned as “hooliganism”. The National Cohesion and Integration Commission has deployed monitors to counties in which it anticipates further trouble, and recent polls suggest that election-related violence is Kenyans' single-biggest worry for 2017.
Even if Cambridge Analytica's contribution only helps Jubilee party to sway a few thousand votes, given how close the election will be (according to some polls), this edge may well be decisive.
Not knowing or not caring?
Does Cambridge Analytica not know this, or does not care?
Even if the company does not have the power to influence significant numbers of voters the mere perception that Cambridge Analytica’s service, which relies on in-depth profiling of voters and which presumably includes ethnicity – will add more wood to the pyre. Currently there are no that once an extensive database of citizens’ sensitive information is collected it won’t be abused.
In light of this, Privacy International has written to Cambridge Analytica with questions about how the company assessed the risk of their work in Kenya and how it will ensure that Kenyans’ personal data will be protected.
|PI letter to Cambridge Analytica re Kenya .pdf||158.7 KB|