|Originally published in Inside the Influence Industry's Personal Data: Political Persuasion - How it works by Tactical Tech's Data & Politics team.|
When you tweet your opinion about Brexit or Trump, you probably don't expect the content of your tweet to become part of an analysis of public opinion on the topic. But the emerging field of digital listening draws from your individual interactions on social media - along with that of others - and analyses their content to assess the feeling of individual potential voters and the overall public mood. Although the majority of conclusions from digital listening studies about public political opinions are published by academic researchers or NGOs, the method is also sold as a service as a way of gathering intelligence for political campaigns.
Digital listening is an umbrella term for monitoring and analysing what someone does or says on social media platforms such as Facebook and Twitter. Both the behaviour (retweeting, liking, sharing an image or commenting on a post) and the content (hashtags, tweets, posts and comments) are analysed. Companies who offer these services are able to measure which topics are being discussed among users at a given time or to monitor the sentiment of the content, such as whether people feel positively or negatively towards a candidate. One such company, Bakamo.Social, which works with governments, NGOs and political parties, explains the services it offers through digital listening on their website:
'Bakamo go way beyond keywords and sentiment. From the gritty detail we derive broad themes that attract and motivate people to join the conversation. We understand the full social discourse, chart consumer journeys, define segments based on needs, identify factors that catalyze product choice, and more. Through their authentic voices, you get real, nuanced, and unexpected insight into consumer behavior.'
Traditionally, political strategists and campaigns use polls, calls and canvassing to ascertain voters' opinions and to take the political temperature. Digital listening technology allows them to do the same kind of analysis as these conventional tools, but far more quickly, with fewer resources, and to study larger groups of people.
When that analysis is combined with other datasets, such as lists of the users' followers or their location, digital listening can measure the public opinion of a targeted group of people, making it a valuable tool for political candidates and campaigns.
Digital listening involves two components that automation has accelerated and scaled up: data acquisition and data analysis. First, data is gathered through software called scrapers, from social media posts, tweets connected to a hashtag, or from certain sets of people on Twitter or content from comments on Facebook posts. Data about behaviours is also gathered which help show 'engagements' on the platform such as retweets on Twitter or likes on Facebook. These interactions can be ranked as positive or negative engagements with a topic.
Next, this data is analysed using algorithms to infer different pieces of information, such as whether a tweet demonstrates a positive or a negative sentiment, by analysing the words and context in which they appear. Much of this analysis builds on recent advances in natural language processing (NLP), a kind of artificial intelligence that specialises in looking at large bodies of text. NLP is programmed not only to recognise positive and negative sentiments of certain words, or the linguistic context for the sentiment of a message, but also to develop new rules as it performs more and more analysis, making it 'smarter' over time.
Digital listening technologies, rather than replacing older tools, are usually used in conjunction with them. For example, one traditional polling organisation, YouGov, has been gathering public opinion through emails or other online methods for political parties, governments and private companies. In 2018, YouGov purchased an AI company called Portent.IO for its digital listening capabilities to complement their work. Portent.IO, rebranded as YouGov Signal, carries out text and behaviour analysis on Twitter to 'distill key insights around overall engagement, opinion and market efficacy'. This is used to help understand how well any company, campaign or individual is viewed by the public, which can be helpful for politicians to understand how they can improve their status in the eyes of potential voters.
Data collected from social media by political campaigns can achieve a number of outcomes:
↘ The data can help campaigns understand if a candidate or issue is perceived positively or negatively, what sort of language they are associated with, and how much they are talked about.
↘ The data can show what issues people care about by analysing the most talked about topics or trending hashtags throughout an election cycle.
↘ The software can also help identify political influencers by looking at who has the furthest reach on social media and who has positive sentiment towards a political campaign.
NUVI, a social listening tool developed by Brickfish, offers services devoted to politics. The headline of their politics page reads: 'Understand what is important to your voters at any given moment. Monitor the trends and concepts that your voters are sharing and stay on top of emerging ideas'. Among their services they offer to help political actors to 'stay on top of what voters are thinking', 'create lists of your influencers and detractors, and be alerted when they are talking about certain topics', 'measure sentiment on specific issues' and 'visualise real-time conversation data in dashboards you can access on your mobile device'.
The company Crimson Hexagon, which also specialises in digital listening, offers details on how they gather insights about current political discussions from social media. For example, they recently analysed which candidates have been talked about most when announcing their run for US President in 2020. According to their information, in late 2018, 'when Kamala Harris announced she was running on Jan. 21, there were 191k tweets about her candidacy. Bernie Sanders also generated 191k tweets when he announced he was entering the race on Feb. 19'. They also measured public sentiment about the elections and found that many voters were 'sad' about critical issues such as climate change and immigration; some people had 'fear' and some had 'joy' surrounding the new candidate announcements.
Another company, Ossalabs, markets their Election Impact tool specifically for political campaigns. Ossalabs advertise that through their tools they can help politicians 'prepare for public questions by keeping your fingers on the pulse of constituents' top of mind issues', 'discover and respond to small crises that impact electorate decisions before they become too large', 'anticipate impending attacks from your opponents' and 'understand which talking points and topics are resonating'.
Digital listening can be used to accomplish a variety of goals. OssaLabs, based in Virginia, explains how campaigns can use their services in this screenshot taken from the website. The company’s name comes from Ossa, the Greek goddess of both rumour and fame.
Source: ‘Political Campaigns’, OssaLabs, accessed 12 March 2019.
These three companies show the variety of types of opinions that can be measured to inform candidates about whether their talking points are working or if they should adjust them according to public sentiment.
In Taiwan: The firm AutoPolitic worked on the mayoral campaign for Taipei in 2014. AutoPolitic, in their words, 'crawls and transforms social media data into actionable intelligence'. A first-time candidate running for Mayor of Taipei, Ko Wen-je worked with AutoPolitic. For Dr. Ko's campaign, the company measured public sentiment to understand 'what topics the public cared about (and why), who the influencers are (so they can engage them) and what topics the influencers are most interested in'. The firm generated a list of activities and ranked them based on their predictions about how much engagement they would get based on past actions online. They concluded that Dr. Ko should engage with young people through activities including tattoos, street dancing, basketball and riding bicycles. Dr. Ko followed this advice and his visit to a tattoo parlour was considered successful, as it was shared widely on social media platforms.
In India: Germin8 Social Intelligence is an Indian company that has provided digital listening research in politics. Germin8's 'social command centre' is online software that monitors 'social conversations' such as debates on Twitter or public Facebook pages. They published analysis of these conversations in the run-up to the 2014 elections and published the results online, available for anyone to use. The results showed that the Bharatiya Janata Party (BJP) had a more positive message focusing on hope, whereas the Aam Aadmi Party had a critical approach that focused on issues such as corruption. A Germin8 spokesperson said how this probably impacted the success of the campaign, as a positive message is more appealing to first-time voters. This shows how digital listening can provide information that can be integrated into strategies for future campaigns.
Digital listening companies frame their methods as measuring what is said in 'public'. There is little transparency as to the full extent of this monitoring of 'public' space. Some organisations are explicit that they are on Twitter to monitor Twitter behaviour; others are not. This makes it difficult to know definitively whether you are being 'listened' to. You can, however, assume that by talking in a public online space, such as Twitter, or if you have a public Facebook account or talk in public Facebook groups about politics, your data could be collected and used in ways outlined above.
↘ Digital listening can circumvent some of the problems associated with conventional opinion-gathering, such as self-censorship and the Hawthorne effect (the effect by which subjects may behave differently when they are aware that they are being observed).
↘ Digital listening allows campaigns to assess and measure the opinions and sentiments of much broader and larger groups of people than traditional methods of polling and surveys.
↘ Digital listening focuses on behaviour instead of aspiration or attitude - it provides 'unfiltered' opinions.
↘ Users rarely provide or are asked for explicit consent to be part of digital listening analyses, but companies justify it by the collection of data from 'public' spaces. Bakamo.Social's slogan, 'Insights without asking' suggests that this lack of consent can be seen as an advantage.
↘ Though digital listening service providers suggest that 'sentiment is pretty simple to understand. It's just a feeling or emotion, an attitude or opinion', opinion is not 'simple' to gather. Rather, digital listening focuses on present tense as a way of predicting what people want or will want in the future - which is not necessarily a reliable method.
↘ Further, though digital listening might help reach or measure different sets of groups to those who are surveyed through traditional techniques, it is limited only to those people engaging in political discussions through social media online and therefore gives a limited perspective.
Author: Amber Macintyre