The Problem of Voting From a Technological Lens

In every presidential election to date, no more than 45% of the eligible voting population has cast their vote. As an individual, it’s unlikely that your vote will swing the election. However, if everyone thought this way, the candidate that wins might not best represent the wishes of the people, and people that do vote might have overrepresented power in government. This presents an interesting challenge, how do you represent the peoples’ mandate when not even half the people actually chose a person to support? To answer this question, we’ll look at how conceptual voting in tech applications works and examine how to influence the candidates in an election to better represent your views from that lens.

Voting isn’t a unique problem. To ascertain the wishes of the people, to understand their preferences and desires, to create policies that better serve the public; these are all goals that many modern-day algorithms attempt to solve. Every time we make a Google search, we cast a ballot for which website should be ranked at the top. Whenever we ‘like’ a photo in a Facebook newsfeed, we cast a ballot for which photos our friends should also see. When we purchase an item on Amazon, we cast a ballot for which items should be recommended to others. Sharing our preferences with these apps is actually a problem very similar to the one of voting in this country.

Let’s take a closer look at how this might work by examining a mock shoe recommendation algorithm on Amazon. In many of these algorithms, there’s usually an option to ‘not share your data.’ This basically means that you’ll get the same experience as others, but data from your activity won’t be used to make other peoples’ experiences better. (Not sharing your data would be similar to choosing not to vote) Say you recently bought a pair of Nikes, but chose not to share your data with the recommendation algorithm. As a result, it may underestimate the demand and popularity of Nike shoes. If everyone that bought Nikes did this, then the recommendation algorithm might not suggest Nikes at all. Of course, realistically, since Nike is such a popular brand, it’s unlikely that your data will make much of a difference in the recommendation algorithm much like how people perceive their vote doesn’t matter. In tech, this gives people the perception that sharing their data is just an unnecessary hindrance to their privacy for no benefit.

But this comparison still doesn’t answer the key question. Why do fewer than half the eligible voters in the US actually vote?

The major differences between how Amazon recommends shoes and how we vote are the granularity of the data and the default choice.

When you create a recommendation algorithm, you can use a variety of “input features”. These features contain information about you like your age, location, or purchase history and are used to rank the possibilities that appear. If you’re from a certain area, you might like items branded with your local sports team. Similarly, for the other features, they might adjust the ranking up or down as well. What’s important though is that typically in these recommendation systems there are so many features that it’s possible for each person has their own unique combination of features that defines them. There are so many features that the algorithm likely has never seen a person with that representation before, but must make a decision about what to show you. Google must give you results for all the searches that you do, even if you are one of the 15% of people that search for something never searched before every day. This level of granularity is something that can only be handled in tech.

In the offline world, it’s almost impossible to see this level of granularity catering to personal preferences. You would likely need a personal team of people to understand your thought process, your ideals, and your choices. In the world of politics, you would need to be able to say not just whether you were pro-life or pro-choice, but under precisely what conditions an abortion would be considered acceptable. Would it be acceptable if the mother’s life was in danger? Would it be acceptable a day before the expected delivery? Where would you draw the line? There’s a host of other controversial issues where this level of granularity is frankly impossible to measure or find even though drawing a hard line is often necessary when writing and enforcing laws.

The other major difference between voting in tech and politics is that voting is the “default option”. As the default, you don’t need to spend extra time and energy to explain your preferences, and it’s also much harder to misrepresent those same preferences. For example, when you make a Google search, that search is saved in your preferences as your search history unless you choose to delete it. Similarly, as you browse links and add to your history, the algorithm gets a better sense of who you are and what you care about, without you needing to explicitly tell it what to do or instructing it to write down certain bits of data. This means that you don’t need to spend extra time or effort to vote. The algorithm takes care of that for you.

However, in some cases, having a process with extra time or money as a component creates an even better algorithm. For instance, Google likely tracks the amount of time you spend looking at a certain YouTube video to recommend others similar to it since the time you spent looking at a video you don’t want to watch would be a high cost for you to pay (unless, of course, you weren’t in front of the screen, but that’s a whole separate matter).

Indirect methods like a poll are also much less common in apps since you can more easily misrepresent your preferences. For example, if I enjoy window shopping, I might say that I want to see more of a certain brand on Amazon, even if I don’t have intentions to ever buy that brand. Instead, it’d be much easier for Amazon to track preferences through my purchase history because I likely wouldn’t spend so much money to modify the recommendations.

In politics, unfortunately, the default is not voting, and there are no other direct methods for measuring someone’s beliefs as purchases would for a shopping recommendation algorithm. Since the range of issues covers everything the government handles, the scope is just so broad that any one issue is unlikely to cover it. There’s also the possibility that people feel strongly about issues they don’t personally experience. For example, it’d be hard to measure abortions as a proxy for support for or against abortions.

The closest indirect proxies are close interest groups and political parties. Close interest groups like a police union may endorse a particular candidate or convince them to adopt certain nuanced positions that would better their members. Political parties similarly represent a certain set of beliefs and by ascribing to that party, you lend your credence to it and give it the power to shift a candidate’s position. For example, Joe Biden inherited a more liberal Democratic platform despite his personal background as closer to a moderate Democrat.

So again, if voting can’t show your detailed position on the issues, and there’s a high cost to doing so, with a high degree of unreliability in the real world, what’s the point of voting? How can you make sure your voice is heard?

I believe that this long chain of problems and differences between consumer choice as evaluated in tech and in political voting shows why less than half the population votes. But, I still believe that having your voice heard is important. It’s just much harder.

As with Joe Biden above, supporting other candidates that are closer to your position and having your desired candidate pull the platform in a certain direction is a great way to have your voice heard. This is slightly limited by our two-party system, but there is no limit on the number of outside groups you could also support. You can support certain environmental groups if you believe strongly in fighting climate change. You could also support police unions or groups for police reform. The availability of different groups out there allows you to represent the granularity of your beliefs and amplify your voice on nuanced positions as much as possible without tech. Most importantly, to sway the candidate closer to your position, and support any of the groups above, you’ll need to vote.

Local voting information and drop box locations



Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store
Alan Liu

Alan Liu

SWE @ Nuro | Formerly Facebook/Google | Yale ’18 |