Can technology predict the US election outcomes?
The prediction of election results is a rather interesting exercise in the analysis of the sociopolitical makeup of a country. Over the several months leading up to the formal day of the election, pollsters take up the unenviable – and rather arduous task of placing thousands of voters under the microscope and parsing data from telephone surveys and past voting trends to determine possible outcomes from the ensuing election. But one mustnot place all their bets in the process – polling isn’t an exact science. Far from it, in fact.
Historically, election predictions have often turned out to as wrong as right. In fact, if we are to move back to the lead-up to the 2016 US presidential election, most reports declared Hillary Clinton a clear winner nationally, with tighter races reported in certain states like Michigan, Wisconsin and Pennsylvania. Donald Trump, however, did manage to triumph, gathering 304 electoral college seats – much higher than the required 270-seat margin. A report from the American Association for Public Opinion concluded that the gross underestimation of Donald Trump’s support, especially in the upper Midwest, was owing to a lack of high-quality polling data more than anything else.
And, this is where the true magnitude of the challenge comes in. Polling forecasts are usually made up of traditionally collected sources of data, such as from online surveys telephone calls. Unsurprisingly, however, given the fact that the quantity of data collected is comparatively small and rather untrustworthy, one must ponder about whether there is a better way to go about the forecasting process.
Several analytics-based firms, such as Expert.AI, Advanced Symbolics and KCore Analytics have claimed that building algorithms on election dynamics whilst drawing data from social media signals such as tweets, Facebook posts or messages might be a more prudent means of forecasting election results. This follows the now oft-used business trend of gauging consumer sentiment through the use of technology such as artificial intelligence. Although far from perfect, what this process does afford is a much more rounded view of election dynamics. It’s accuracy – especially in the aftermath of the 2020 election – is, however, a different story altogether.
Method Behind the Madness
Polling itself is an act that is fraught with uncertainties owing to the diverse methods involved. Likewise, some of the disparity in the algorithm-driven forecasts can be attributed to methodological differences.
Expert.ai leverages a knowledge graph that identifies named entities – including people, companies, and places – and attempts to model the relationships between them. The company says its system, which attaches 84 emotional labels to hundreds of thousands of posts from Twitter and other networks, semi-automatically weeds out bot-like social accounts. Expert AI’s algorithm ranks the labels on a scale from 1 to 100 (reflecting their intensity) and multiplies this by the number of occurrences per candidate. At the same time, it classifies emotions as either “positive” or “negative” and uses this to create an index that can compare the two candidates.
KCore Analytics, which claims to have used over 1 billion mined tweets to guide its predictions, taps an end-to-end framework to find influencers and hashtags in networks like Twitter. Data is selected according to both content and frequency — ostensibly in real time and excluding bots — which an AI model called AWS-LSTM analyzes for opinion classification, with a claimed accuracy of up to 89.5%.
Advanced Symbolics’ Polly, developed by scientists from the University of Ottawa, gathers a randomized, controlled sample of American voters identified by their posts and conversations on social media.
The challenges in predicting election results with AI are many. One is that the algorithms must be trained to learn different models for the electoral college that coincide with national predictions. Another is that they need to fine-tune their ability to uncover issues important to specific minority groups and regions. The smaller the groups, the harder these are to find. Moreover, none of these models considers the way legal challenges, electors who don’t vote for the candidate they’d pledged to, or other confounders might affect the outcome of a race.