Algorithms Are Getting Better at Identifying Human Behavior

The holiday season has many things associated with it that generates positive emotions in people around the world. Sadly, it is also a time that generates negative feelings which includes suicidal thoughts.  Although humans are getting better in understanding the human psyche, suicide among young adults is the second-leading cause of death according to the U.S. Centers for Disease Control and Prevention.  Fortunately, artificial intelligence is improving in this field as algorithms are getting better at identifying human behavior.

What is an Algorithm?

Let’s first explain what an algorithm is, which is a procedure used to help find the answer to a mathematical problem (finding the greatest common divisor) in a number of finite steps which frequently involves repetition of an operation; in other words, a step-by-step procedure for figuring out the answer to a problem or accomplishing some end by using a computer.  While this may be a lot to take in, just remember that AI has evolved from simply being used to retrieve data to answer factual questions (think of the show Jeopardy) to now being able to identify specific human behaviors.

Source: Medium

How algorithms have been identifying human behavior

Back in 2014, an article by Jennifer Golbeck focused on the type of insights one can make about an individual by analyzing social media accounts.  Considering the amount of data that is shared on various social media sites, the information that some may think is innocuous is revealing; analyzing tens of thousands of other profiles can lead to secrets one never intended on sharing.  The title of the article, Smart People Prefer Curly Fries, shows that using the right algorithms and combining it with meaningful data can result in revealing human behavior/characteristics that even the individual was not aware of.

There was a time that before artificial intelligence and algorithms were around, it was up to humans to figure out how other humans were feeling and what actions may occur as a result such as determining if an individual will want to harm another person or themselves.  When technology advanced to utilizing AI and algorithms that can solve certain problems faster than people could on their own, many presumed that machines would not be better at understanding how humans felt than their fellow humans.  Yet, a previous study suggested that this was not the case.

Writer Olivia Goldhill in 2015 wrote an article that stated an algorithm could predict behavior in people better than humans could and was based on an MIT study that suggested an algorithm could not only predict an individual’s behavior quicker but also to be more reliable than a human could.  This was achieved by a master’s student in MIT’s computer science named Max Kanter and a research scientist at MIT’s computer science and laboratory in artificial intelligence advisor named Kalyan Veeramachaneni.  They created a Data Science Machine that could look for patterns and pick which variables would be the most relevant; their paper regarding the results of the project was presented in Paris at the IEEE Data Science and Advanced Analytics conference late in October of 2015.

Source: LinkedIn

Moving forward, others continued to work on using AI and algorithms to make predictions in human behavior and an article written by Carlos E. Perez in January refers to research that shows how ground breaking techniques are being created. MIT was able to train a system known as Predictive Vision to focus on YouTube videos from programs like Desperate Housewives and The Office whether two individuals will slap a five, hug, shake hands or kiss; they instructed the Deep Learning system using 600 hours of video.  The system predicted an action 43 percent of the time, which comparing to previous algorithms that predicted only 36 percent of the time.

Obviously, algorithms are getting better over time on predicting human behaviors quicker and more accurately than humans, which can prove helpful in a variety of ways.  There was an article in May that suggested a mood-recognizing algorithm might help save people’s relationships. Imagine an app that can be use on a wearable that could keep a couple informed on if there’s a change in their partner’s mood.  Once receiving an alert, the other person would know to tread lightly to avoid needless conflict. Sounds good, right?

The sad news is we’re not there yet; however, new research supports the idea that in the not-too-distant future, an app like this for wearables could be achieved.  Researchers at the University of Southern California were assisted by machine learning methods to construct an algorithm that uses mild changes within an individual’s physiology to determine her or his emotional state; the eventual goal of the researchers is to make an app that couples could try that would lower the amount of conflict in their relationship.

Source: IAP

The future for algorithms identifying human behavior

While having an algorithm that can identify in a relationship the mood change of one partner in advance to avoid unnecessary conflict is nice, will this be the limit to what algorithms will be used for that identifies human behavior?  That would be a resounding no considering where this research started and where its’ focus is today.  Considering how I mentioned in the beginning how high the rate of suicide is in young adults, writer Emily Watz last month posted a piece that identifies the importance of the algorithms in the field of those working to recognize suicidal behavior and to assist before it is too late.

There is a future for this type of technology and it is just a matter of time before the progression of algorithms recognizing elements of human behavior will have its’ greatest impact on humanity.