Identity in an AI World
Identity is the story we tell about who we are — to ourselves and to others. It is built gradually, through experience, reflection, relationships, and the choices we make about how to present ourselves to the world. This process has always been social: we learn who we are partly by seeing how others respond to us, by comparing ourselves to those around us, and by inhabiting the roles and categories our culture makes available. AI is now deeply embedded in all of these processes. It shapes how we look in photos, how others perceive us through algorithmic categories, how we encounter others online, and how we understand our own preferences. The effects on identity are subtle, pervasive, and worth examining carefully.
The Filtered Self
AI-powered photo filters and editing tools can change a face in seconds: smoother skin, larger eyes, slimmer features, different hair color, different lighting. Millions of people now post photos of themselves that have been substantially altered by AI tools before anyone else sees them. On TikTok and Instagram especially, the filtered version becomes the version others know — and sometimes the version the person themselves most identifies with. Researchers in psychology and dermatology have documented a condition informally called Snapchat dysmorphia: people arriving at plastic surgeons not with photos of a celebrity they want to look like, but with AI-filtered photos of themselves that they want to look like permanently. The filter has become the target. This is a specific and striking version of a broader phenomenon: when AI can produce a more attractive version of your face on demand, the unfiltered face begins to feel like a deficiency rather than the baseline reality. This is not simply vanity. It is a distortion of the identity-building process. Healthy identity development requires some acceptance of and integration with your actual appearance, history, and characteristics — not as fixed limitations, but as the real material of who you are. When AI makes it easy to display a substantially different self, the relationship between the presented self and the real self can become genuinely confused.
Sociologist Erving Goffman distinguished between the self we perform for others and the self we experience privately. Some gap between these has always existed and is normal. AI amplifies this gap: the performed self can now be much more polished, optimized, and idealized than the experienced self. When the gap becomes very large, it can generate anxiety, inauthenticity, and disconnection from one's actual life.
Deepfakes — AI-generated video or audio that realistically depicts real people doing or saying things they never did — pose a more extreme identity threat. Anyone with sufficient data about your face and voice can potentially create a convincing fake of you. Deepfakes have been used to create non-consensual intimate imagery (a form of sexual abuse), to spread political disinformation, and to conduct fraud. The fundamental violation is identity theft in a new form: the separation of your likeness from your consent. Beyond individual harm, deepfakes erode a foundational epistemic norm: that video and audio evidence corresponds to something that actually happened. When this norm weakens, it becomes harder to hold people accountable for their actual words and actions, because the defense 'this video is fake' becomes more plausible. Researchers call this the liar's dividend: even if a person did say something damaging, they can more credibly deny it in a world where fakes are known to exist.
Match each AI-identity phenomenon to its clearest real-world consequence.
Terms
Definitions
Drag terms onto their definitions, or click a term then click a definition to match.
Algorithmic Identity and Self-Understanding
Beyond appearance, AI shapes identity through categorization. Every major platform builds a model of you: your interests, demographics, political tendencies, purchasing behavior, emotional patterns. These categories are used to serve you content, target you with advertising, and determine what opportunities you see. In a meaningful sense, the platform's model of you is a kind of identity — a set of characteristics attributed to you by an AI system. The problem is that algorithmic categories are self-reinforcing. If a platform categorizes you as interested in fitness, it serves you more fitness content, which increases your engagement with fitness, which confirms the category. You may become more interested in fitness partly because the algorithm kept directing your attention there. Is that your authentic interest, or a preference the algorithm helped construct? The honest answer is: both, and it is increasingly hard to tell them apart. This matters for identity because people have a right to develop their interests and values through genuine exploration, not just through what a profit-driven algorithm decides to reinforce. The algorithm does not know what is good for you — it knows what you have engaged with, which is not the same thing. Self-knowledge requires some space to encounter things you did not already want, to discover interests you did not already have, to revise your sense of yourself in light of genuine surprises. Algorithms that serve you only what you already are tend to make you more of what you already are — which is a form of stagnation dressed as personalization.
One practical response to algorithmic identity reinforcement is deliberate exposure to what the algorithm would not serve you: following people with different views, reading outside your genre, watching creators from different cultural contexts. This is not about forcing yourself to like things you do not like — it is about keeping your identity-forming process more fully under your own direction.
A teenager's social media algorithm consistently serves them content about entrepreneurship and financial success after they engaged with one video on the topic. Over time, they begin to strongly identify as a future entrepreneur. Which statement best describes the role AI played in this process?
What is the 'liar's dividend' as it relates to deepfake technology?
Map Your Algorithmic Self
- Most platforms will show you their category model of you if you look in privacy or settings. On a platform you use regularly, find the section that shows your inferred interests or ad categories.
- Step 1: List the top 10 categories the platform has assigned to you.
- Step 2: For each category, rate it on two scales:
- a) How accurate does this feel? (1 = completely wrong, 5 = exactly right)
- b) Is this a category you chose deliberately, or did it emerge from habitual behavior? (Chose / Emerged / Mixed)
- Step 3: Identify two categories where you feel the algorithm got you right in an interesting or surprising way.
- Step 4: Identify one category that feels reductive or wrong — where the algorithmic label misses something important about you.
- Step 5: Write a paragraph responding to this question: If this algorithmic profile were the only information someone had about you, what would they understand correctly, and what would they fundamentally miss?
- Discuss: What does the exercise reveal about the difference between behavioral data and genuine self-knowledge?