Tinder is the swindler!

I watched The Tinder Swindler, a documentary film about a con man who met women on Tinder and manipulated them emotionally into financing his lavish lifestyle. Wow, what a documentary! Online dating sites and apps have fundamentally changed the dating landscape, as Finkel et al. argued in their insightful 2012 study about the psychology of online dating.

While watching the documentary, I kept thinking about the inner algorithmic workings of online dating apps, their dark side, and the extent to which data generated by online dating apps can be used in quantitative finance.

Opening the black box of dating algorithms

This 2019 Vox article describes quite well the algorithms behind major dating apps like Tinder and Hinge. Initially, Tinder used the Elo rating system, according to which you (as a user) rise in the ranks based on how many people liked you (swiped right when seeing your picture). However, this rank is weighted based on who the people liking you are. If these people liking you have got a lot of likes themselves, then when they like you your personal score goes up more.

It seems that now Tinder uses more cutting-edge technology after Match Group (Tinder’s parent company) took over Hinge, another famous online dating app. Hinge uses machine learning based on the Gale-Shapley algorithm.

According to this article on The Verge, Hinge’s technology “breaks people down based on who has liked them. It then tries to find patterns in those likes. If people like one person, then they might like another based on who other users also liked once they liked this specific person.”

The dark side of online dating apps: user data extraction

I recently came across the work of Joana Moll, a Barcelona/Berlin based artist studying digitalized capitalism. She gave a presentation on ‘Data extraction, materiality, and agency’ organized by Northampton Contemporary Art, a contemporary art space in Northampton, UK. Moll’s latest work is called Carbolytics and examines the environmental impact of advertising technology (adtech). Importantly, in 2018 Moll together with Berlin-based Tactical Tech published the work called The Dating Brokers: An Autopsy of Online Love.

For the Dating Brokers project, Moll and Tactical Tech purchased 1 million online dating profiles for €136. These dating profiles included almost 5 million pictures, usernames, e-mail addresses, nationality, gender, age and detailed personal information (e.g., sexual orientation, interests, profession, physical features, etc.) of users from several online dating apps. This batch of data was bought from USDate, which is one of many data brokers trading dating profiles to populate new and existing online dating sites.

Moll and Tactical Tech showed that online dating companies such as Plenty of Fish, Tinder, Hinge, OkCupid actively share and sell user information with their parent company Match Group and a myriad of third-party companies. This is a large network of companies extracting data and capitalizing on such data without user’s consent.

Such practices raise major ethical and legal concerns. As Moll and Tactical Tech put it, data from online dating apps travel “far and wide and could potentially be instrumentalised by third-parties for advertising and individualised pricing, but also to restrict your access to health insurance, credit, education and much more.” Furthermore, dating profiles contain “intimate and sensitive data on users which, if made public, might have dramatic effects on the user’s lives.”

In a sense, Tinder is a swindler using deception to obtain our data.

This below is a screenshot of Moll and Tactical Tech’s The Dating Brokers project. Visit, support, and share.

In parallel to Moll’s argument, The Wall Street Journal has recently published this exclusive article about gay-dating app Grindr collecting user locations and selling such data to advertising partners.

It is not just about online dating apps…

This FT article on data brokers by Aliya Ram and Madhumita Murgia is an insightful examination of the data brokerage industry, their data mining practices and technologies. Data brokers collect thousands of data points to build up extensive profiles of individuals and sell related analytics. As the article explains:

Data brokers mine a treasure trove of personal, locational and transactional data to paint a picture of an individual’s life. Tastes in books or music, hobbies, dating preferences, political or religious leanings, and personality traits are all packaged and sold by data brokers to a range of industries, chiefly banks and insurers, retailers, telecoms, media companies and even governments. The European Commission forecasts the data market in Europe could be worth as much as €106.8bn by 2020.

Meanwhile in the US…

One of the largest data marketplaces is Oracle, the computer software company based in California. Oracle owns and works with more than 80 data brokers who funnel in an ocean of data from their own range of sources, including consumer shopping behaviour at brick-and-mortar stores, financial transactions, social media behaviours and demographic information. The company claims to sell data on more than 300m people globally, with 30,000 data attributes per individual, covering “over 80 per cent of the entire US internet population at your fingertips”.

Check this picture below. It is taken from the FT article about data brokers and is scary.

Regulators are now trying to rein in ‘super data brokers’ like Acxiom (now LiveRamp), Oracle, Nielsen, Salesforce, Experian, Equifax, Criteo, Quantcast, Tapad. The objective is to develop a regulatory environment that guarantees data privacy and ethics.

If you want to study an insightful theorization of how data are generated, managed, and assembled into tradable objects, you should read this fantastic paper by Aleksi Aaltonen, Cristina Alaimo and Jannis Kallinikos on ‘data commodities’.

Quantitative trading and alternative data

Do quantitative traders use alternative data from online dating apps? I am not sure, maybe. Evidence is scarce.

I read this great paper by Kristian Bondo Hansen and Christian Borch on alternative data and sentiment analysis. It is a good starting point to explore the issue further, to the extent that any data from online dating apps would belong to the umbrella category of ‘alternative data.’

We know that the big data industry is growing and quantitative hedge funds are big fans of alternative data in their endless search for above-average ‘alpha’ returns.

This FT article by Lindsay Fortado, Robin Wigglesworth and Kara Scannell does a good job of explaining the main types of alternative data being used in finance, the gold rush of alternative data, and mentions the problems concerning privacy.

See also this other blog post of mine on alternative data. This graph by AlternativeData about the overall alternative data space is quite useful to map where online dating apps would be situated.

Privacy is a big problem because big data sellers (e.g., Thinknum, Qandl) often fail to hide personal information. This other FT articles by Robin Wigglesworth delves deeper into the privacy issue. It has a crucial passage:

“The vendors claim to strip out all the personal information, but we occasionally find phone numbers, zip codes and so on,” said Matthew Granade, chief market intelligence officer at Steven Cohen’s Point72. “It’s a big enough deal that we have a couple of full-time tech people wash the data ourselves.” The head of another major hedge fund said that even when personal information had been scrubbed from a data set, it was far too easy to restore. “We were shocked at how easy it was to de-anonymise the data,” he said. “It took one of my analysts 30 minutes to discover someone who was probably having an affair.”

The question of how to derive useful information from alternative data while respecting privacy is leading to developments in so-called ‘synthetic data’. The latter involve “funnelling real-world data through a noise-adding algorithm to construct a new data set. The resulting data set captures the statistical features of the original information without being a giveaway replica.” See this FT article by Anjana Ahuja here.

Anyway, back to the main question: do quant traders use data from dating apps? Possibly. These apps offer lots of data that are useful to make predictions in the stock of a dating company that is publicly listed, like Bumble (NASDAQ: BMBL) or a listed parent company such as Match Group (NASDAQ: MTCH), as well as the overall online dating industry.

How do we extract data from dating apps? This blog and its follow-up post are fascinating. The author gathered a data set from Tinder by using bots swiping about 300 profiles a day in Germany. Data were taken through an unofficial Tinder API. The author did not share the data set to avoid legal issues.

To be continued when I gather more evidence…