Earlier I came up with an idea of 16: Interview for a city leader.

The other day during a water cooler chatting, a coworker suggested that we could use AI to select president candidates. This is actually a very good idea.

Why using Machine Learning to select Candidates is a good idea?

In US, if someone is interested in participating presidential election, he or she has to hire a group of people and then go through a long dragging campaign journey, such as going to different cities and giving talks, sending out flyers, and meeting with all kind of people. Well, at least that was the story in Man of the Year. Hopefully the candidate is meeting with the real people working hard and trying to live, not just the riches and drink champagne together.

If you think about it, it is a very costly process. Therefore, a candidate usually needs a sponsor. The issue of having sponsors is: nothing is really free. The candidate then is assumed to have to pay back from another way. Then how is this a fair and trust-able game? The candidate is actually just working for the companies or individuals who sponsor the process, not the people who are looking for a better leader.

That’s why the so-called election is probably just an illusion.

However, if candidates are selected by computer AI, things could be much more interesting, and fair.

Let’s arbitrarily list some data points:

• Experience: the person probably should have been involving in social work;
• Presentation: the person probably should be a good presenter;
• Public Speaking: in addition, this person probably should be comfortable anytime in public;
• Accomplishments: what has this person done so far;
• Social contribution: like what this person has written, journals, tweets, comments in social network, etc.
• Integrity: credit score, tax records, etc.

Those data can be gathered from Internet, public videos or references, as well as some financial records.

To have a training model, we can first use past presidents, let’s say we have 10 to 20 most recognized as good presidents, like we acknowledge their work and accomplishments when they were presidents. We use the data like: what they said, what decisions they made, what they wrote, and so no as training data to establish a model.

Next, this ML application will be running for a year. It (probably) monitors all senior level people who work in the government and analyze their behaviors, gathers the real-life data and feeds to the training model, slowly filter out people who do not seem to qualify. In the end, after a year, the application will create a list of people (say under 10) who are believed to qualify as presidential candidates.

This software saves the time for candidates to run campaign. In order to be selected, one who is interested in becoming a candidate must do all the work in public to generate more data points for the software to pick up. This motivates and encourages the potential candidates to really work, not just keep talking and talking.

I feel this might even greatly increase the efficiency of  government.

Risks

Of course there will a lot of risks to implement this idea.

First, for any system, there will be some people come up with another system to game the original system. Eventually, people may realize how the software works, in terms of, saying what words in what situation will help you cheat the selection algorithm.

Second, security. This system must be secure enough so no one can hack into it to manipulate the selection process or the analysis result.

Finally, privacy. To have a fair selection process, people have to expose a lot of private information, such as tax records. I am not sure as a president, if it is legal to make those data public. But on a second thought, if there is nothing to hide, for example, I have never received any unknown deposit from a foreign country, why do I need to worry about it?

Conclusion

I truly feel using machine learning to help select presidential candidates is a promising idea, even though it come with some challenges. Maybe a research lab should consider running a long-term research project or do some POC of it.