Last week, TechRepublic and Unanimous A.I. conducted a 'swarm' to gauge attitudes towards candidates and learn what voters want from leaders, in terms of tech and business. Here are the results.
With the Republican and Democratic party conventions fast approaching, TechRepublic is using a new tool--swarm AI--to make assessments about attitudes towards the candidates and issues in tech and business that matter to voters.
What is it, exactly? Swarm AI is a software platform, developed by Unanimous A.I., that harnesses the power of a crowd to make decisions. It isn't crowdsourcing or polling--instead, it assembles a group online, and makes real-time, group decisions. For instance, the group will be presented with a number of answers for a given question, like "Who will win best male lead at the Oscars?" As soon as the timer goes off, each individual will then pull their "magnet" to their answer. Then, through an algorithm, the group magnet moves towards an answer--if you want to visualize it, imagine a virtual Ouiji board.
Since January, TechRepublic has covered swarm AI as a new tool for predictions. We challenged a swarm to predict the Kentucky Derby in May--and it nailed the superfecta (top four horses, in order), beating 540-1 odds. We weren't as lucky with the Preakness or the Belmont, but the results were still close. The Preakness predictions outperformed the experts, and the results of the Belmont showed something interesting: after our swarm was conducted, the official odds changed to match our picks exactly.
While there are many traditional ways of using big data, social analytics, polling, and experts to make educated forecasts about the outcomes of a political election--which we have been covering in Election Tech--the AI algorithm in the swarm has been incredibly successful, often outperforming traditional methods. While the swarm picks for the Preakness and Belmont weren't quite as successful, the Preakness predictions still outperformed experts, and the Belmont predictions
Thus, TechRepublic and Unanimous A.I. teamed up last week to hold our first of three election swarms, days ahead of the conventions. The swarm was composed of a randomly-selected group, assembled through Amazon Mechanical Turk. There were between 76-86 people who answered each of TechRepublic's questions, all voting-age US citizens.
Unfortunately when we originally conducted the swarm last week, the political divisions within the group made it difficult for the swarm to reach a clear consensus on many of our original questions. So this week, we reworded the questions and held another swarm session. Most of our questions were aimed at how technology and business issues would be affected by each of the candidates. The swarm participants could choose from "no change," "increase a little," "decrease a little," "increase a lot," and "decrease a lot." Here's what they told us:
- How will unemployment change in a Clinton presidency? No change
- How will unemployment change in a Trump presidency? Increase a lot
- How will the US' Cyber Defense change in a Clinton presidency? No change
- How will the US' Cyber Defense change in a Trump presidency? No change
- How will innovation change in a Clinton presidency? No change
- How will innovation change in a Trump presidency? No change
We also couldn't resist asking about the swarm's picks for VP candidates.
- Who is likely to be chosen for the Vice Presidential nominee for the Republicans? Newt Gingrich (options included: Chris Christie, Joni Ernst, Mike Flynn, Newt Gingrich Mike Pence, Jeff Sessions)
- Who is likely to be chosen for the Vice Presidential nominee for the Democrats? Elizabeth Warren (options included: Cory Booker, Julian Castro, Tim Kaine, James Stavridis, Bernie Sanders, Elizabeth Warren)
A major takeaway from the swarm is that, while it does not expect unemployment to change during a Clinton presidency, it does believe a Trump presidency will "increase [unemployment] a lot." The findings could, however, be skewed by the fact that the swarm was a younger group, leaning Democratic. A quarter of the swarm was between 18-25 years old. Forty-two percent were between 26-35; 22% between 36-45, 7% between 46-55, and 4% over 55. The gender breakdown was 54% male, 46% female. And 25% Republican, 47% Democratic, and 28% independent or "other."
This was the first of three swarms. TechRepublic will also conduct a swarm post-convention, before the debates. We will hold the final convention after the debates and before the election.
The conventions will likely give the candidates a chance to connect with their respective bases, which may have an impact on the sentiments of the swarm post-conventions.
Have your own questions you'd like the swarm to answer after the conventions? Let us know in the comments below. And use #ElectionSwarm to follow the conversation on Twitter and Facebook.
- TechRepublic's 'swarm AI' predicts 2016 Belmont winners (TechRepublic)
- TechRepublic's 'swarm AI' predicts the Preakness (TechRepublic)
- Can artificial swarm intelligence predict the Oscars? Here are its picks (TechRepublic)
- 7 trends for artificial intelligence in 2016: 'Like 2015 on steroids' (TechRepublic)
- Q&A: A powerful look at the future of AI, from its epicenter at Carnegie Mellon (TechRepublic)