Serendipity is in vogue. Once a niche concept, research on it has been increasing exponentially the last 50 years. However, the fascination with serendipity isn’t confined to the ivory towers; it is also seeping into the public sphere, lending its name to everything from a cheesy romcom and a fresh Rituals product line, to gift shops and consultancy firms, and even lingerie and latex stores. As Merton and Barber (2011) noted, associated with its popularization “The vogue word” more and more “becomes a vague word.”

Some argue that the vagueness of serendipity is necessary to protect its inherent "magic." While I find this sentiment romantically appealing in theory, it proves problematic in scientific practice; especially in the field of recommender systems, where my research is situated.

Serendipity in recommender systems

When I started studying serendipity in recommender systems, I had no idea what this mysterious concept actually meant. Hoping to find some clarity, I dove into the literature; only to emerge more confused than before. To start with, there are so many definitions out there; almost as many as there are papers about serendipity. Everybody has their own interpretation of which components are really core: is it about feeling surprised or feeling resonance? Is it about encountering something unusual or encountering something relevant? The only thing that scholars agree upon is that it is a slippery concept to grasp.

As a consequence, approaches to operationalize serendipity differ dramatically. For instance, a popular way to measure serendipity is by calculating the difference between a serendipitous algorithm and a primitive baseline (e.g., recommendation based on popularity), assuming that a bigger difference equals more serendipity. Others operationalize it through a metric termed Average Distance Time which measures the gap between when a system recommends an item and when a user would have naturally stumbled upon it (Wang et al., 2019). Still others bypass the approximations and directly ask users whether “The recommendation was a pleasant surprise” (Wang et al., 2020). 

It gets even more confusing. Some of these metrics are not just different—they are diametrically opposed. For instance, Shani and Gunawardana (2011) operationalize serendipity as how different the recommended items are compared to the user’s profile; the bigger the difference, the more serendipity. Conversely, Pastukhov et al. (2022) stress the importance of relevance, which is measured as similarity to the user’s profile. We are left with a problem here: different studies measure fundamentally different use experiences. Yet, they arrive at the exact same conclusion: that their system increased serendipity. 

It is a challenging problem to solve since intuitively all these operationalizations, how different they may be, actually make sense. They are expressions of the slippery nature of serendipity. But how to handle this slippery nature? Since serendipity is, at its core, an experience, we decided to consult the real experts: the users. Drawing inspiration from Makri et al.’s (2014) "serendipity stories," we set out to collect real-world accounts of discovery through recommendations, from Spotify Radio to Instagram Reels, to find out what serendipity looks like in the wild.

Three ingredients of experienced serendpity

We found that serendipitous experiences are characterized by three interconnected components: they all entail unintentionally encountering something fortuitous, refreshing, and enriching. Hooray, we added another definition to the already confusingly long list of definitions! However, the real contribution here is that we find that these three components can be expressed in a variety of ways. For example, an experience can be enriching because you encountered something super relevant for solving a problem you were struggling with, but it can also be enriching because it sparks a new interest. 

These different expressions of the main components result in what we call different flavors of serendipity. In our paper, we describe some of these flavors more in depth. For example, taste-reincarnating serendipity: the unexpected rediscovery of something you used to like back in the days. One participant illustrated this nicely when reminiscing about a TikTok trend using Taylor Swift’s Love Story: “I remembered I liked that song, so I decided to look her up again. That’s how I started listening to her again.” This encounter not only led him to rediscover her music, but it actually sparked a deepened connection, resulting in him buying her LPs and even traveling to her highly popular concerts. 

Another tasty flavor is what we call resonating serendipity. This entails fortuitously encountering something refreshing that really resonates with you on an emotional level. For example, one of our participants recalled that Spotify uncovered a latent need for a specific genre of music that he was really craving at that moment: “The last two months I had this itch. The music I listened to doesn’t do it anymore; I am getting bored of it. Should I listen to some Psytrance? No, that’s not it. Should I listen to downtempo? No, also not it. Searching, searching, searching until I encounter this post of Behemoth on Instagram: ‘Ah I haven’t listened to black metal for a while!’ and indeed, that was my itch.” 

Lastly, for dessert, we present taste-broadening serendipity. This occurs when a user encounters an item that is unusual and at the same time intriguing enough to pull them beyond their comfort zone. One participant experienced this through a Zalando ad for discounted moccasin shoes he admitted he "would never have bought on my own.” He reflected: “It’s strange because I never specifically looked for those shoes, but I ended up buying them and they pleasantly surprised me.” What began as an unexpected purchase eventually reshaped his entire aesthetic, transforming his style from casual sneakers and sweaters to a more refined "casual chic" wardrobe of shirts and moccasins.

Why this menu of serendipity flavors matters

Realizing that serendipity comes in many flavors helps resolve its vagueness; it implies that we should be specific about which flavor we are studying. At the same time, our conceptual framework respects the many ways serendipity can manifest in our lives. Rather than contradicting each other, scholars were describing the same concept, just savoring different parts of it. So, we end with a conclusion that perhaps we knew all along: serendipity is a flavorful concept.

Author bio

Brett Binst is a PhD researcher at imec-SMIT, Vrije Universiteit Brussel. His research revolves around serendipity in recommender systems: How does it express itself? How can we stimulate it? How can we measure it?

Further reading on this topic

Makri, S., Blandford, A., Woods, M., Sharples, S., & Maxwell, D. (2014). “Making my own luck”: Serendipity strategies and how to support them in digital information environments: Strategies for “Seeking” Serendipity and How to Support Them in Digital Information Environments. Journal of the Association for Information Science and Technology, 65(11), 2179–2194. https://doi.org/10.1002/asi.23200  

Merton, R. K., & Barber, E. (2011). The travels and adventures of serendipity. In The Travels and Adventures of Serendipity. Princeton University Press. 

Smets, A. (2025). Intended, afforded, and experienced serendipity: Overcoming the paradox of artificial serendipity. Ethics and Information Technology, 27(3), 33. https://doi.org/10.1007/s10676-025-09841-6 

Wang, C.-D., Deng, Z.-H., Lai, J.-H., & Yu, P. S. (2019). Serendipitous Recommendation in E-Commerce Using Innovator-Based Collaborative Filtering. IEEE Transactions on Cybernetics, 49(7), 2678–2692. IEEE Transactions on Cybernetics. https://doi.org/10.1109/TCYB.2018.2841924 

Wang, N., Chen, L., & Yang, Y. (2020). The Impacts of Item Features and User Characteristics on Users’ Perceived Serendipity of Recommendations. Proceedings of the 28th ACM Conference on User Modeling, Adaptation and Personalization, 266–274. https://doi.org/10.1145/3340631.3394863