By skyforbes 
This blog post summarizes the paper “Not just ‘For You’: How the Algorithmic Crystal Mediates Communication and Identity Work on TikTok’s FYP” by Zoë Natalia Cullen, Angela Y. Lee, Brenna Davidson, Jeffrey T. Hancock, & Nicole B. Ellison. This paper will be presented at the 28th ACM Conference on Computer-Supported Cooperative Work and Social Computing. Read the full paper here.
“The mind is a strange machine which can combine the materials offered to it in the most astonishing ways, but without materials from the external world it is powerless…” — Bertrand Russell
Every day, millions of people open apps like TikTok, Instagram, or YouTube and scroll through an endless stream of content that can, at times, feel eerily tailored to them. This might feel familiar: think about that funny video that perfectly captures your sense of humor, the recipe that matches your dietary restrictions, or the “hot take” that does not actually feel provocative at all but is instead aligned precisely with the way in which you see the world.
Of course, this is neither coincidence nor is it magic. Behind the scenes of these popular platforms, sophisticated algorithms are at work: building detailed profiles of your interests, values, desires, fears, and personality traits based on every tap, pause, like, comment, and share.
These algorithmic systems have proven remarkably effective at capturing users’ personalities and preferences. Many people feel their personalized feeds accurately reflect who they are, including the unique, dynamic, and multifaceted aspects of their identities and interests.
Scholars have described this dynamic through frameworks such as the Algorithmic Crystal, developed by my co-authors, which conceptualizes personalized algorithms as “reflective surfaces” that produce such accurate reflections of us that they can sometimes shape how we see ourselves and others. You can read more about the algorithmic crystal framework here.
However, identity extends beyond the individual. We also carry social identities, that is, identities shaped by our roles, relationships, and group belonging. These social identities are central to how we understand ourselves and connect with others.
In light of this, we wanted to know: How do people use the videos on their TikTok FYP — shaped by their own individual behaviors and interests — to express who they are and strengthen their relationships with others?
To address this question, we conducted a study with 27 regular TikTok users, employing two complementary methods (see Figure 1).
- In-depth Semi-structured interviews — We asked participants to discuss their TikTok use, experiences with and perceptions of their FYP, and their sharing practices
- Live observation sessions — Participants shared their screens over Zoom as they scrolled through their personal FYPs, allowing us to observe in real-time how they engaged with content and interacted with others
For complete details on our methodology and analysis: read our full paper from CSCW 2025.
Turning Personalized Content Into Social Communication
Our key takeaway: People don’t just passively scroll through their personalized feeds to consume content aligned with their individual interests. They actively use algorithmically recommended videos as a form of social communication. This took three primary forms:
- “This is me” — sharing content that reflects their identity, interests, or current emotional state
- “This is you” — sending videos that remind them of a specific person
- “This is us” — sharing content that represents shared experiences or group identity
Beyond recognizing themselves in their algorithms, our participants also saw their close social ties and shared relationship dynamics reflected in their feeds. The algorithm quietly picked up on these patterns and surfaced content that enabled them to communicate who they are both individually and collectively (see Figure 2).
Perhaps more intriguingly, some participants noted a feedback loop at work: the more they shared particular videos with close friends, the more the algorithm surfaced similar content, creating a steady stream of new opportunities to connect.
At first glance these micro-interactions might seem trivial, but they serve as small, subtle, bids for connection; our participants described how such gestures allowed them to nurture and maintain relationships across distance and time. You might think of it as the digital equivalent of seeing something in a store window that reminds you of someone, snapping a photo, and sending it to them — except now a personalized algorithm is doing the window shopping for you.
And yet, while artificial intelligence increasingly facilitates these exchanges (indeed, tallying clicks and views and shares to predict what might reflect and resonate with you or a friend), what ultimately transforms this sterile stream of data and gives it meaning is not necessarily the math behind the algorithms but the profoundly human impulse to share.
In turn, each shared video carries with it the weight of recognition, a quiet affirmation of being seen and remembered. A digital tap on the shoulder that says, “Hi, this made me think of you.”
A New Kind of Digital Identity: The Algorithmically-Networked Self
Such findings invite us to rethink how people develop their sense of self in digital spaces. When we talk about digital identity, we mean the version of ourselves that takes shape online through our activity, connections, and the digital traces we leave behind.
Just like in everyday, offline life, where we are often seen as someone’s friend, partner, sibling, neighbor, or coworker, in digital spaces we build and present identities through the things we share, the people we follow, and the posts we interact with.
Previously, researchers identified two main ways this digital identity formed:
- The Networked Self (Traditional Social Network Sites)
Here, identity forms through direct interactions with people you know. Think Facebook or Instagram, where likes, comments, and conversations with your circle of friends may shape how you are seen and, in turn, come to see yourself. - The Algorithmized Self (Algorithm-Based Platforms)
Here, identity is shaped in isolation, based on what algorithms learn from your clicks, watches, and skips. This has sometimes been described as the “lonely algorithm problem” — a feed that knows you intimately, but leaves you socially disconnected.
Our findings do not align neatly with either of these categories, but instead point to a hybrid dynamic, one that blends the relational intent of the networked self with the curatorial mechanisms of the algorithmized self, which we refer to as the algorithmically-networked self.
Even though TikTok’s FYP doesn’t explicitly connect you to friends the way Facebook does, people still found their close relationships reflected in the algorithm’s recommendations. Videos reminded them of friends, captured shared experiences or inside jokes — and users leveraged such moments to reach out and communicate something about themselves.
This creates a reinforcing cycle: When users send algorithmically-curated content to friends, they’re teaching the algorithm something about their social world, gradually fine-tuning recommendations to match not just individual preferences, but the shared tastes of users’ social networks. In effect, the old saying “you are who you surround yourself with” becomes “your algorithm is who you share with.”
How it works:
- The algorithm curates content based on your individual behavior.
- You recognize social meaning in that content (“this is so my friend”).
- You share it, using the algorithm’s picks as tools for social connection.
- This creates a feedback loop where individual preferences and relationship dynamics mutually reinforce each other.
Our research highlights how personalized algorithms have become more than private reflections of the self, but socially responsive channels through which people communicate and maintain their connections.
The algorithmically-networked self complicates the idea that algorithmic personalization isolates us. Instead, our findings highlight how people creatively employ AI-curated media to signal who they are, recognize others, and affirm shared bonds.
For more details and full description of the research context, methods, findings, and discussions, please refer to our full paper, archived and presented at the 28th ACM SIGCHI Conference on Computer-Supported Cooperative Work & Social Computing (CSCW) held at Bergen, Norway on October 18–22, 2025.