top of page

Swiping Right for Artificial Intelligence

Today, roughly 20 percent of all relationships and 17 percent of all marriages begin online. With about 2,000 dating apps operating in the US and almost 8,000 total operating worldwide, online dating has quickly become a multi-billion-dollar industry with most projections predicting at least $3 billion in revenue in the US alone for 2020. Over the years, companies within the digital space have sought to remain relevant by integrating new technology to push the boundaries on efficiency — the online dating industry is no different. One subset of technologies that has proven indispensable for these issues is Artificial Intelligence (AI) and Machine Learning (ML). Both have quickly become no stranger to dating apps around the world. Nowadays, apps like Badoo, Betterhalf, eHarmony, OkCupid, Hinge, Tinder, and have all heavily integrated AI/ML into their platforms to handle a variety of issues ranging from user compatibility to fraud detection and user moderation.

User Compatibility

Picture online shopping, where a user starts with thousands of potential products before selecting some filters to help narrow their search. The individual clicks the option to sort by “price low to high” and then begins to scroll through their filtered results. Originally, this defined the matching process behind online dating, an impersonal filtration process with no method of differentiating results. The goal of online dating is to match users with other compatible users until a perfect match is made, and at first the task of finding a perfect match came down to user trial and error. The incorporation of AI/ML led to a new process that’s dynamic, streamlined, and truly personal to each user. Rather than solely looking at filtering criteria, these technologies have been deployed to gather and utilize much more in-depth data on each individual user. Past profile preferences, previous profiles that were viewed or interacted with, and conversations conducted within the application are all examples of data being continuously mined in real-time. Algorithms are designed to handle this influx of incoming data and dynamically generate models that analyze past behavior in order to predict future preference. The more someone uses one of these apps, the more effective these algorithms become as they can develop a more comprehensive view of a consumer’s personal taste.

Before vs. After Integrating ML Into Dating Apps

Some examples of this technology in action include:

  • eHarmony utilizing Natural Language Processing (NLP) principles such as keyword extraction and sentiment analysis. These tools enable the application to analyze user text data, such as in-app conversations and responses to questionnaires, to identify areas of interest or topics of conversation. To take it one step further, the app is then able to detect an opinion (typically on a scale of positive to negative) towards these topics of conversation in order to match the user in question with other like-minded individuals.

  • Badoo deploying Deep Learning Neural Networks to analyze images and facial features of profiles a user previously liked in order to develop a holistic impression of their physical preferences. The app is then able to show more profiles that tend to share these similar features and image content.

  • Hinge incorporating the Gale-Shapley algorithm, an algorithm developed in 1962 by mathematicians David Gale and Lloyd Shapley which proved a solution to The Stable Marriage Problem. In the context of Hinge’s platform, this method seeks to find a stability point amongst all potential user pairs: where there is not one person in a pair who prefers someone from a different pair who also prefers them in return.

  • DNA Romance is collecting DNA samples and preference data in combination with AI algorithms to match users by genes supposedly linked to attraction. Although the scientific research behind the validity of this matching process is still pending, the end goal for this app and apps alike is to have a process that is dictated by biological compatibility.

Fraud Detection and User Moderation

With a vast market for online dating comes massive potential to bring happiness to its millions of users. However, there is a dark side to online dating as well where scammers seek to abuse these platforms for their own unethical intentions. Some common forms of fraud include emotional manipulation, extortion, blackmail, and catfishing. This behavior going undetected can lead to victims with emotional damage, financial scarring, and sometimes consequences which are much worse. AI/ML-based methods have become crucial in providing security to deter and eventually stop scammers from taking advantage of innocent victims on these platforms.

Supervised ML methods using rich labeled data on both fraudulent and legitimate profiles are frequently deployed to identify the probability of profiles being fraudulent (regression) and to classify both fraudulent and legitimate users (classification). Regression techniques being used include logistic regression and bagged decision tree/random forest regression. Classification techniques include k-nearest neighbors, support vector machines, bagged decision tree/random forest classification, and logistic regression with a threshold.

In addition, unsupervised ML methods, such as cluster analysis (k-means clustering, hierarchical clustering), are incorporated heavily to help identify new behavioral patterns that can uncover previously unseen methods of fraud.

Even NLP is used in a similar manner to its applications in user compatibility in order to both discover and understand a user’s true intentions. User moderation is required to determine where behavior may not necessarily be fraudulent but is certainly still inappropriate. Effective user moderation methods bear many similarities to those typically seen in fraud detection, but also include the addition of utilizing ML for image classification in order to identify and flag inappropriate picture content.

Finding love can be difficult. Both in-person and online applications settings provide their own set of unique obstacles. Since its inception, the online dating world has broken the boundaries of matchmaking by setting out to creating a virtual platform for people to connect and foster real relationships. By embracing change, being unafraid of emerging technology, and integrating AI/ML into their processes, dating apps around the world have successfully given people real opportunities at finding their ideal match.


About the Author

Amit Gattadahalli is an Associate at Boulevard with a focus in data science. He recently graduated from the University of Maryland College Park with a B.S. in Mathematics and Statistics. Since 2018, his work within the consulting industry has largely concentrated on machine learning, custom data science-centric algorithm development, data visualization, and general software development.


Join us as we explore the future of business through the lens of machine learning & AI, data-driven analytics, cloud computing, and operational transformation.





bottom of page