top of page
  • Writer's pictureJohn Ennis

Eye on AI - April 15th, 2022

Welcome to Aigora's "Eye on AI" series, where we round up exciting news at the intersection of consumer science and artificial intelligence!


Do humans and AI think alike? Our post this week addresses that question and more by looking into a new technique that compares how well AI’s reasoning matches that of a human, then shifts gears to address AI-assisted dating’s digital future.


New Technique Aim’s to Address AI’s Explainability Problem

It’s no secret that AI is a wonderful pattern-recognising technology. What’s less obvious is how and why complex AI models make their decisions. This leads to what Industry insiders refer to as the black box problem (i.e. an AI in which the inputs and operations are not visible to the user or another interested party). Because AI is often infused with biases, either because of biases within the data the model is trained on or biases inherent in the developers that created algorithms for the models, it’s significant for users of it to understand how and why decisions are being made to cultivate trust in the technology.

“In machine learning, understanding why a model makes certain decisions is often just as important as whether those decisions are correct,” writes SciTechDaily contributor Adam Zewe. “For instance, a machine-learning model might correctly predict that a skin lesion is cancerous, but it could have done so using an unrelated blip on a clinical photo.”

To address this issue, researchers are searching for new techniques to help understand why AI decisions are made. Last week, one such technique, Shared Interest, was unveiled by researchers at MIT and IBM and shows not only why an AI model’s decisions were being made, but also whether those decisions matched the reasoning of a human.

“... Shared Interest could help a user easily uncover concerning trends in a model’s decision-making — for example, perhaps the model often becomes confused by distracting, irrelevant features, like background objects in photos,” continues Zewe. “Aggregating these insights could help the user quickly and quantitatively determine whether a model is trustworthy and ready to be deployed in a real-world situation.”

The beauty of Shared Interest is that it allows users to aggregate, sort, and rank individual explanations from a machine-learning model and rapidly analyze its behavior. It works by comparing saliency methods to ground-truth data, then searching for alignment (or misalignment) in how well the model’s recommendations match human reasoning.

A good example is the first case study listed in the article, in which Shared Interest was used to help a dermatologist determine if he should trust a machine-learning model that he was considering using to diagnose cancer in patients using skin lesions photographs. Shared Interest shared examples of the model’s correct and incorrect predictions to the dermatologist, then let him decide whether the model was accurate enough to implement (ultimately, the model made too many recommendations based on image artifacts rather than actual liaisons, and the dermatologist decided it wasn’t accurate enough for implementation).

Explainability in AI is one of its biggest hurdles. If people don’t know they can trust a model’s decisions, they’re less likely to use it. Innovations like Shared Interest help give AI the credibility it needs for wider adoption, proof that the predictions being made are accurate and align with human reasoning. However, like AI models themselves, Shared Interest’s effectiveness is only as good as the methods it is based upon. If those techniques contain bias or are inaccurate, Shared Interest will inherit those limitations.

The Future of Dating in the Age of the Metaverse

Changing gears, let’s conclude by looking at how dating is evolving in the age of the metaverse. According to futurist Bernard Marr’s recent article “Looking For Love In The Metaverse: Dating In The Age Of AI and VR,” AI-assisted dating apps are progressing rapidly to meet the demands of our digital future.

“Dating apps are big business, with over 1,500 dating apps and websites on the market today. Experts predict that the market for dating applications will reach $9.2 billion by 2025,” writes Marr. “People worldwide are looking for ways to connect – and now it seems like smarter technology to enable those connections might be right around the corner.”

Tinder’s CEO explained how he believes that in the not-so-distant future, a dating app may be so good at predicting our ideal match that instead of users having to swipe through endless potential connections, they can simply log on and have the app recommend a potential partner nearby, explain why he or she would make a great match, then recommend a restaurant or concert it recognizes you’re both interested in. Dating possibilities go beyond that through the use of VR and AR. At this very moment, companies like FireFlare Games are working to develop dating experiences aimed at connecting people virtually before meeting in person.

“FireFlare will include a selection of settings (including a bar and a forest walk) where users can talk, meet, and potentially create lasting relationships,” writes Marr. “Planet Theta’s website says: ‘These VR dating experiences are an opportunity for you to explore each other's personalities and passions. We are putting human connection at the forefront of dating again. Step foot onto Planet Theta and join the evolution.’”

There are always those who prefer to meet the old-fashioned way. For those that struggle to find love in a world that’s becoming increasingly more socially distant, AI-assisted dating could help connect them with that special someone they’ve been searching for.

Other News


That's it for now. If you'd like to receive email updates from Aigora, including weekly video recaps of our blog activity, click on the button below to join our email list. Thanks for stopping by!

bottom of page