Brett Binst, Ulysse Maes, Martijn C. Willemsen, Annelien Smets
10 min
Abstract
Research on how people experience music emphasizes the importance of exploration and diversity in listening. However, music recommender systems struggle with facilitating exploration. Even when music recommender systems are able to recommend something valuable to users that is outside their typical preferences, it still remains difficult to spark their interest. This paper presents a user study examining the efficacy of immersive and informative introductions in stimulating interest in songs that are beyond one's usual preferences, an experience called Taste-Broadening Serendipity. We uncover two important mechanisms behind the effect of introductions: transportation and cognitive elaboration. Our findings indicate that transportation (i.e., being absorbed into a narrative world) is the strongest predictor of Taste-Broadening Serendipity, while cognitive elaboration (i.e., learning something new about the artist or social context in which the music emerged) has a weaker effect but is easier to stimulate. We propose that song introductions can play an important role in facilitating exploration and increasing diversity of listening on music streaming platforms.
Sam: It captured both: real actions like likes, dislikes, skips, and total play time, plus quick surveys on what people felt during each song. This let them check if intros changed how folks saw the music's trickiness or if they could handle it.
Alex: Right, so past systems nudged exploration but didn't fully spark interest in the wilder genres. What about that SIRUP model? How does it fit?
Sam: SIRUP applies the idea that interest grows when something's new but not overwhelming. Think of it like spotting a new video game level that's fresh but matches your skills. It first checks if a song is fresh compared to your listening history using linked data paths—like tracing family trees of similar tracks—and then gauges if your past plays show you can handle variety. This study builds on that by focusing on intros to tweak those feelings directly.
Alex: So SIRUP spots potential surprises, but intros make them stick by adjusting how complex or doable the song seems. And for the factual intros, they leaned on this cognitive mastering idea?
Sam: Yes. Cognitive mastering sees listening to tricky music—like jazz or opera—as solving a puzzle: you hunt for meaning in the sounds until it clicks and feels rewarding. Without background, it stays confusing; the intros supply quick facts about the artist's life or song's roots, linking it to what you already know so the puzzle solves faster.
Alex: That tracks—it's like getting hints for a tough video game level. Makes the unfamiliar less scary.
Alex: So with that web app capturing real listens and feelings, what did the actual setup look like to make sure the comparisons were fair?
Sam: Everyone who joined tried all three intro types—no intro, story one, fact one—but the order got shuffled randomly for each person, so no one type always came first or last. They only counted trials where people confirmed the song felt outside their usual style. The flow had quick sign-up with basics like age and country, then listening to three songs from one genre with randomized intros, rating feelings right after each, and a wrap-up survey on wanting more from that style.
Alex: Okay, so same songs per genre but mixed order, and only unfamiliar ones—that controls for personal biases nicely. How many people went through it, and how diverse were they?
Sam: They started with 475 participants, mostly from a site called Prolific for broad reach, plus some from their network, then filtered to 350 solid sessions. Ages averaged 36, about half women, half men, from 37 countries.
Alex: Right, sounds solid for spotting patterns. And those ratings after each song—what exactly did they measure to link intros to interest?
Sam: After each song, people rated on simple agree-disagree scales how much they felt pulled into a story world, gained new insights linking to what they knew, sensed they could grasp it, or found it deep yet not shallow. These grouped into reliable sets tapping those pathways. For the main outcome, they made a yes-no score: song differs from usual, sparked interest, plus one surprise or fresh-view item—all high marks. That binary captured true broadening cleanly.
Alex: Huh, so not just likes, but these feeling clusters. How did they crunch the numbers to see if intros really drove that yes?
Sam: They used multilevel stats that handle grouped data—like ratings nested in people and songs—modeling paths from intros through feelings to the outcome. One check ran a Bayesian logistic setup for the yes-no odds, showing immersive intros nearly doubled the chance over none, informative ones boosted it more than twice.
Alex: Wait—so both worked, but song-to-song differences mattered. Does that mean the story pull shone on some tracks more than facts?
Sam: Yes, the data shows clear variation across individual songs—for instance, some heavy metal tracks hit over 75% success with immersive intros, while others stayed under 20% regardless. That highlights how the intro's fit to the specific song influences results.
Alex: Huh, so it's not just the type of intro, but how well it lands on that track. What about when intros felt off—did they backfire?
Sam: Exactly. They grouped user ratings of intro quality into low, medium, and high. Low-quality informative ones actually lowered success rates compared to no intro at all, while high-quality versions strongly boosted odds—immersive about two and a half times higher, informative nearly four times. This underscores that execution matters.
Alex: Right, poor facts could make it worse, like bad hints confusing a puzzle. So how did intros change those core feelings of trickiness and graspability?
Sam: Listeners rated two key senses after songs: first, how much depth or layers they perceived in the music—like noticing intricate rhythms or emotions that reward attention. Second, their confidence in unpacking it—like feeling equipped to follow along without frustration. The study found success peaked when both were high, around 87%. Researchers label these *perceived complexity* and *coping potential*.
Alex: Okay, so intros lifted both to make unfamiliar songs feel rich but approachable. But was that the full story, or did other paths explain it better?
Sam: They mapped the connections using a statistical method that traces chains of influence—like arrows showing how one feeling leads to another. It revealed immersive intros worked entirely through boosting coping potential, while informative ones routed through both complexity and coping. Robustness checks confirmed these links.
Alex: Wait, so story intros mainly built handling confidence, facts added depth sense too. Did those feelings fully account for the gains?
Sam: Not quite—the deeper model showed transportation and cognitive elaboration feeding those, but with direct lines to success. Stripping out coping potential, transportation emerged as the top driver overall, especially immersive intros. Informative ones leaned more on elaboration directly. In total, these paths explained a notable share of broadened interest.
Alex: So transportation carries the main load for stories pulling you in. And beyond immediate sparks, did it lead to wanting more from the genre?
Sam: Yes, a targeted analysis compared songs liked versus those sparking success in a session. More successes strongly predicted stronger desire to explore the genre, outpacing likes.
Alex: So overall, these intros seem like a solid nudge toward broader listening, especially since more sparks predicted wanting to dive into whole genres.
Sam: Yes, and the paths clarify why: transportation pulled hardest through handling confidence, while elaboration offered steadier gains via both depth and grasp. Informative intros proved more reliable overall, as immersive ones varied more—fragile if the story misses.
Alex: Right, so no perfect fix—depends on the track. But with execution key, what limits did they flag in pulling this off at scale?
Sam: A few notable ones. Effects shifted a lot by song and intro quality—poor ones even hurt more than nothing. The AI voices sometimes eroded trust, as users questioned the info's source. Plus, the lab setup with required ratings felt less natural than casual streaming.
Alex: Huh, those make sense—AI tone could turn people off, and forcing surveys changes the vibe. Still, does the paper point to fixes or next steps?
Sam: It suggests tailoring by listener type—like story fans for immersive, fact seekers for informative—and testing real apps or human voices. Long-term, platforms could watch for listening ruts and push "taste-breaker" playlists with these intros, auto-made via language models for scale.
Alex: Okay, so a practical path forward, balanced by those real-world hurdles. This study lays out clear mechanisms and evidence for why intros work, pointing recommender systems toward better discovery.
Sam: Precisely. It shows contextual previews meaningfully lift interest in the unfamiliar through distinct psychological routes, with room to refine for everyday use.
Alex: That's a grounded look at how song intros can spark broader listening. Thanks for joining us on ResearchPod.