I just read a story about a 19 year old Stanford student who has developed a legal bot that can help people appeal parking tickets. The app only works in the UK, but so far it has successfully appealed over $3 million in tickets.
Joshua Browder, the student, programmed his robot based on a conversation algorithm. It uses keywords, pronouns, and word order to understand the user’s issue. He says that the more people use the robot, the more intelligent it becomes. Its algorithm can quickly analyze large amounts of data while improving itself in the process.
Apparently, Browder’s isn’t even the first lawyer bot. The startup Acadmx’s bot creates perfectly formatted legal briefs. The company Lex Machina does data mining on judges’ records and makes predictions on what they will do in the future.
After reading the article, and looking at the types of questions the program asks the user (check out the image above), it seems like such a tool would be helpful in guiding someone through the process of choosing a political candidate to vote for.
For example, it may ask questions such as the following:
- Do you believe the wealthy should be required to share their wealth with the less fortunate?
- Do you believe abortion is wrong under all scenarios?
- Do you believe that health care is a right for all people?
- Do you believe immigrants should be able to attain citizenship?
- Do you believe that someone under the age of 50 could be an effective President?
- Do you think unions are bad for our country?
- Do you think there is a need for greater gun control?
That’s just a small sample of the types of questions that could be asked, and it seems that to be effective that there should be at least several dozen questions. Many of the questions could be multi-part, branching off in certain directions depending on how the user answers a question.
Behind the scenes there needs to be a lot of data about where the various candidates stand on the issues; perhaps the candidates themselves would be the ones supplying the data. Their answers could be vetted by an independent group (if there is such a thing), like the League of Women Voters, before being uploaded into the database.
Users would also be able to weight their answers based on which topics are most important to them. For example, before answering a question, or series of questions on abortion, the program could ask “On a scale of 1-10, how important is the issue of abortion to you?”
The bot would then take a person’s answers, and generate an overall score for each candidate. The one with the highest score would be the one whose political views are most closely aligned with yours on the issues that are most important to you.
Since it is highly unlikely that any one candidate has views that are perfectly aligned with any voter, there recommendations would be based off the relative weighting that is given to the various topics.
For example, a candidate might be strongly in favor of gun control and strongly in favor of abortion. A voter might be strongly in favor of gun control, but strongly against abortion. If the voter weighted the gun control issue as being more important to him, then this candidate would do OK. If on the other hand, the voter ranked abortion as one of his top issues, this candidate would not come out as favorable.
With dozens (if not hundreds) of questions spread across all of the key issues, the voter would get a comprehensive analysis of which candidate is most closely aligned with his beliefs.
Again, the voter may still disagree with some of the views of the top candidate recommended, but all things considered that candidate should be their choice.
I realize there’s a lot of things that need to be worked out here, but for something I just thought of an hour ago, I think it has some potential.
Any Stanford computer science students want to go in on this with me?