As it is election season going on in full fervour, it would be only fair to discuss predictions and results. Exit polls have traditionally been conducted on voters after they cast their vote at the polling booth, or over phone calls for those who choose remote ballot options. For the most part, it has been a reliable way of gauging the pulse of the electorate, but we've seen misses too, most recently in the US presidential election 2020. So it may be worth asking, how good are these exit polls? And do we have a better alternative?

Turns out, the answer to this polling problem lies in AI, much like most other problems of a technologically driven world. Poll prediction is basically a function of data and sentiment analysis, and these are tasks that machine learning is particularly good at performing. And we have precedent too – AI was able to correctly predict the outcome of the Brexit vote in 2016. 

AI and ML can leverage the tons of user-generated data on the internet that is indicative of their opinions, sentiments and political leanings. Many experts agree that it is a more accurate way for predicting results, and it will keep getting better because every subsequent election will be a new data point for the algorithm to learn from and improve itself.

There are a few AI pollsters that are doing notable work. KCore Analytics is a company that is in the business of forecasting behavioural trends from real-time social media. Their algorithm had predicted Biden’s victory by a narrow margin in the electoral college. Advanced Symbolic Inc. had deployed its business analytics tool called Polly for political forecasting in the US election, though it couldn’t get it quite right this time. Expert.AI used sentiment analysis to label the opinions being expressed by people on social media and, as per a report by Venture Beat, had projected a 50.2% chance of Democratic victory.

The challenges faced by AI pollsters are similar to those of traditional ones, however. The biggest drawback encountered is the lack of representation of people from rural and remote areas who have been difficult to reach first by phone and now through digital channels. But the comparative advantage that AI enjoys keeps getting bigger every year, and it is only a matter of time until we can expect near-accurate predictions by AI-enabled pollsters.

Image from Pexels