Artificial Intelligence Catching Gang Members? Researchers Say Possible

Are AI enabled algorithms ready to help police do their work?

Many people would argue that the police has been a little less than perfect as far as preventing and stopping crime is concerned.

And that might be the reason why some researchers believe that the answer could lie in the use of Artificial Intelligence in order to fight crime and rid cities around the world of “bad hombres”.

Of course, you won’t be seeing any Robocop styled killing machines out in the open hunting down criminals any time soon. What researchers are trying to look into is Twitter. Yes, Twitter, researchers believe, can become a useful tool for Artificial Intelligence engines to identify and then scout potential gang members.

Obviously, researchers are assuming the fact that potential gang members will actually use Twitter. There is no doubt about the fact that social media feeds around the internet contain mountains of precious personal information.

From daily quibbles to comments about movies and music along with plans for a night out with friends. So should one really wonder as to why the police would want to have a way which would enable them to mine the data that is available with different social media sites?

Hypothetically speaking, the data provided to the police by any or every social media site could prove immensely useful in presenting insights into where criminals could hang out or strike next.

But as far as Artificial Intelligence is concerned, researchers are now talking about a level that goes beyond stopping crime by knowing the whereabouts of criminals in different parts of the world or a given location.

What the researchers are looking into is the possibility of these digital relics, taken in their entirety, collectively being able to convey something or anything deeper about individuals who are using these social media websites.

Some researchers even want to study the possibility of Artificial Intelligence being able to predict who an individual really is just by analyzing that individual’s digital footprint.

As indicated before, a number of researchers now believe that in the very near future, complex computer algorithms which will be trained to work with this kind of information may be able to make crucial decisions about certain individuals or suspects.

Readers who follow Artificial Intelligence news and how Artificial Intelligence is affecting every industry that is even remotely connected with a computer and an internet connection, might recall a recent example where researchers working from the  Ohio Center of Excellence in Knowledge-enabled Computing (Kno.e.sis) at Wright State University came up with some pretty interesting results through a published study.

The researchers published a paper to arXiv preprint server which stated that the team working on the project was able to devise a new deep learning Artificial Intelligence computer algorithm which could identify various common gang members based on nothing more than their Twitter posts.

Not only that, but the new Artificial Intelligence enabled computer algorithm could successfully ascertain street gang members with a pretty decent accuracy of 77 percent.

Of course, that does not justify the rest of the 23 percent where the Artificial Intelligence enabled computer algorithm is likely to make a mistake rather than identify the correct individual.

When Motherboard contacted one of these experts, the expert revealed that the current state of this technology (Artificial Intelligence enabled computer algorithm to identify potential criminals) had some rather severe shortcomings which rendered the Artificial Intelligence enabled computer algorithm more or less useless since it was likely to do more harm than good.

The expert further continued and said that it was certainly within the realm of possibility that the Artificial Intelligence enabled computer algorithm could determine an individual as a gang member based solely on that individual’s likeness for rap music or specific emojis that the individual had the habit of using frequently while communicating to his peers on social media websites.

Current AI technology is several handicapped at best.

In other words, the current Artificial Intelligence computer algorithm technology needed to make use of more number of criteria in order to judge someone as a gang member.

Perhaps it is appropriate at this time to mention that the paper published by team Kno.e.sis has not been peer reviewed yet.

Regardless, the paper further states that the Artificial Intelligence enabled computer algorithm developed by the team is trained through the use of a database of “gang members” Twitter accounts and profiles.

These Twitter profiles were identified as such by the researchers through previous studies where members of the team subjectively interpreted tweets of “gang member” Twitter profiles and then based their findings on those subjective interpretations.

Researchers also managed to find out that the distinguishing characteristics of the gang member Twitter profiles that were included in the study were,

  • Use of offensive words, such as the n-word
  • Rough talk with/about other people
  • Generous use of emojis related to services such as gas pumps, which can sometimes translate to weed
  • A fondness for rap music

Researchers combined these characteristics together and then used them as the basis for the team’s description of “self-identification” of suspected gang members on the social media website, Twitter.

It should also be pointed out that the Artificial Intelligence enabled computer algorithm did not take into account information regarding the location of the “gang members” or any city or area specific usage terms or even hashtags.

Researchers reasoned that local terms specific to an area or a city changed within a very short amount of time and that rendered them useless for the purpose of the study.

We can’t say much about the researchers but right now, it sounds like like if someone likes to upload rap music to YouTube or share rap music links from YouTube on his/her Twitter account AND does not shy away from using the n-word AND is a little too habitual with using specific emojis AND likes to put up images related to subjects such as money or presumably other items that researchers associated with gangs, THEN the Artificial Intelligence enabled computer algorithm might peg that special someone as a gang member.

Amit Sheth, who is the study author and an executive director of Kno.e.sis, wrote in an email that no one of these ‘features’ would lead the company’s Artificial Intelligence enabled computer algorithm to assert one individual as a gang member.

He further wrote that the team of researchers, exploited observations such as these: doing rap does not imply that you are a gang member but listening to gangster rap music increases a possibility that one has a gang association.

Researchers are using training databases to train their AI enabled computer algorithms.

The researchers involved in developing the Artificial Intelligence enabled computer algorithm, back up this claim by citing a paper, published by another group of researchers who argue that social media preserves rap’s upshot of “keeping it real”.

The authors of the other published paper also acknowledge that rap indeed helped shape “the rebellious, assertive voice of predominantly urban youth.”

With that said, it is also true that the “other” paper does not explicitly state that if someone listens to rap music or likes to post links on social media that are related to rap music has an increased chance of being a member of a gang.

Sheth was quick to clarify this fact when, in an email conversation with Motherboard, he said that no single ‘feature’ such as listening or liking rap music would be used to identify any individual as a gang member, however, a combination of many of these features each using known affinity to gang activity certain lead to the good result the researchers reported in the published paper.

There is no doubt about the fact that gang members certainly use social media websites to communicate with each other and to other online users but isn’t that the case with almost everyone else?

Readers should also know that the people working at the various police departments actually know this fact as well.

However, it should be noted that any attempt made by researchers, through the use of Artificial Intelligence, to tag someone as a potential gang member based on the sole input of social media activity should be understood by also taking into account other highly controversial Artificial Intelligence gang-member finding initiatives such as the California’s CalGang database.

It might be hard to believe but the aforementioned CalGang database was put together by researchers through the use of perfunctory indicators such as the individual’s clothing preferences along with tattoos and types of tattoos.

As a result of these measures, the database is said to contain the names of many people who are in reality, not gang members. Of course, these unfortunate people are treated by the police as such anyway.

As far as we can tell, there are a couple of serious problems related to Kno.e.sis’s Artificial Intelligence enabled computer algorithm which could, in all likelihood, land researchers in a similar situation to California’s CalGang database.

The first and perhaps the most prominent of these problems (speaking from a technical perspective of course) is that the training methods used to improve the Artificial Intelligence enabled computer algorithm were not reliable. Researchers used a training database that consisted of gang members and their particulars to train the Artificial Intelligence enabled computer algorithm.

The only problem was that these “gang member” names were not verified in person by the researchers and any third-party participant in the research team. From what has been revealed so far, the research team basically collected user profiles that made use of specific content related to language, music, and emojis and then proceeded to make the conclusion that the profiles that matched these criteria were actually gang members in real life.

For now, AI must be complimented with human oversight to identify potential gang members.

Needless to say, that the Artificial Intelligence enabled computer algorithm was trained to learn from the aforementioned database and hence was evolved to select user profiles that fit an identical template.

As indicated earlier, there is no guarantee that the training methods used by the researchers to train the Artificial Intelligence enabled computer algorithm , actually resulted in the database set containing any gang members, to begin with, it stands to reason that any individual that is tagged by the Artificial Intelligence enabled computer algorithm as a potential gang member, is not to be taken at face value.

Proponents of Artificial Intelligence will say whatever they want to, but common sense dictates that the Artificial Intelligence enabled computer algorithm is fundamentally flawed because it bases its classification on a default set of assumptions regarding the kind of people who like to listen to a specific music or make use of certain words.

Perhaps we should also mention the fact that considering people, who like to listen to a specific set of rap music or like to talk in a specific manner, as gang members is a widespread assumption made by people who the society only knows as racists.

It doesn’t matter if these assumptions are made online or offline, they should be considered as entering the realm of racism.

This is also the reason why a lot of emerging issues regarding the development of Artificial Intelligence are related to the problem of machines being trained on collected data that is prejudiced in nature.

And when these Artificial Intelligence enabled machines are put to work, they tend to come up with results that are very similar to the ones produced by humans as far as prejudices are concerned.

Desmond Patton, who is a professor at Columbia University’s School of Social Work, recently said in an interview that social media was a useful tool for understanding the environmental and situational factors that influenced how and why people engaged in aggressive communication.

Desmond had done extensive research in using computer algorithms to identify youth that is suspected to be at risk of committing crimes or engage in an unconventional behavior.

Desmond Patton was also the lead author on the research paper that was cited by Amit Sheth which we mentioned in the article before.

In an interview given to Motherboard, Patton stated his opinion and said, “And that might be an individual who is gang-involved, or individuals who mimic gang-like behavior because they live in neighbourhoods where gang violence has a large presence and they need to engage in survival strategies to stay safe,”

He further added that they may talk tough or use language that was similar to gang-involved youth to present themselves as being tough online. Patton also noted that while specific types of rap music, like “drill”, were commonly associated with gang members, having a preference or linking rap music was something entirely different since there were probably millions of people on earth, from every professional field who did the same every day.

Patton ended his interview by saying that it was true that youth were not inherently violent and often times the things communicated online were a function of broader issues that were unfolding and that it was important that researchers though less about labels and trying to identify groups and instead focused their energies on things that shaped behavior.

Zohair A. Zohair is currently a content crafter at Security Gladiators and has been involved in the technology industry for more than a decade. He is an engineer by training and, naturally, likes to help people solve their tech related problems. When he is not writing, he can usually be found practicing his free-kicks in the ground beside his house.
Leave a Comment

This site uses Akismet to reduce spam. Learn how your comment data is processed.