A team of computer scientists–including a Moroccan professor– from New York University has devised a methodology to evade the typical search bubble suggestibility approach utilized by most current search engines, with an algorithm called Pyrorank, which draws from the natural world in mimicking interactions in an ecosystem.
The study appears in a newly released volume Advances in Swarm Intelligence (Springer Nature), entitled Pyrorank: A Novel Nature-Inspired Algorithm to Promote Diversity in Recommender Systems.
Pyrorank is a novel bio-inspired re-ranking algorithm designed to improve recommendation diversity. It is inspired by the positive effects of pyrodiversity in nature and is optimized to increase user-based diversity and mitigate the systemic bias that traditional recommender system models learn from the data.
“When it comes to inspiration for solutions to computer science problems, nature is the perfect place to look,” said clinical associate professor at NYU’s Courant Institute of Mathematical Sciences and a co-creator of the algorithm, Anasse Bari.
Recommendation systems are algorithms that utilize data to advise or promote items or options to customers based on the users’ prior purchases, search history, and demographics. Recommendation systems are used by Google, Netflix, and Spotify, among others. However, because these factors place consumers in filter bubbles, they skew search results.
Bari, who oversees the Predictive Analytics and AI Research Lab at the Courant Institute, notes that the traditional manner in which recommendation systems work is by basing recommendations on the notion of similarity. He provides the example that an Apple product user will accordingly receive more Apple product recommendations in future searches.
In startling ways, the shortcomings and inefficiencies of the current algorithm have come to light. Political partisans, for example, are primarily guided to news content that aligns with their pre-existing views. Obviously, this serves only to reinforce their political stance, closing their minds off to objectivity.
Even more disturbingly, though, recommender systems have even propagated self-harm videos for susceptible individuals.
In testing the viability of the algorithm, the researchers compared the search outcomes generated by the Pyrorank add-on with those from traditional recommender systems by using three large datasets: MovieLens, which offers user-generated movie ratings; Good Books; and Goodreads. (The latter two store readers’ book ratings). They then conducted a series of experiments to determine which systems created a greater diversity of recommended content, while at the same time maintaining consistency with the aims of the core recommendations.
Overall, the systems utilizing Pyrorank produced suggestions that were more diversified than those produced by the ones which were already in place, thus proving its usefulness in dismantling the filter bubble design.