cs.HCOct 7, 2026
SaveAI-Mediated Self: How HCI Defines and Relates to the Self
Organizations: Cornell University Ithaca, New York, USA
Abstract
How might AI alter how we understand and experience the self? This scoping review analyzes 102 papers to examine how the self is defined in the field of human-computer interaction (HCI), how AI-self relationships are conceptualized, and what risks emerge when AI becomes entangled with selfhood. Our synthesis makes three contributions. First, we define AI-mediated self as a conceptual umbrella that connects dispersed work across education, workplace, health, and creative practices. Second, we consolidate six framings of the self with four domains of ethical risk-agency/autonomy, identity/authorship, relational capacity, and meaning-making-into a conceptual map that provides a reusable vocabulary across contexts. Third, we introduce the Inclusion of AI-Self framework, which situates AI-self relationships along a spectrum of proximity. Together, these contributions position selfhood as a central design space in HCI.
Figures & tables
Figure 1 . The Proximity Framework for AI–Self Relations , surfaced in our scoping review of 102 HCI papers from 2013 to 2025. Increasing overlap between the two circles marks increasing proximity between self and AI, from Separated to Merged. Four pairs of circles labeled Self and AI with increasing overlap, from Separated to Merged.
| Criterion | Included | Excluded | Boundary case |
| Publication scope | Peer-reviewed archival research articles published in the selected HCI sources by July 10, 2025. Empirical, conceptual, and design-oriented papers were eligible. | Non-archival items, papers outside the selected sources, and records published after the search date. | A technical or conceptual paper was eligible only if it also met the AI-involvement and self-engagement criteria. |
| AI involvement | A user-facing system retrieved through our AI-related terms that performed a conversational, generative, predictive, classificatory, or personalized function in direct interaction with people. | AI used only for backend automation, infrastructure, system optimization, or algorithmic performance evaluation without direct relevance to people’s interaction or experience. | A rule-based chatbot could be included when it was treated as an intelligent interactive agent and participated directly in a self-related practice. |
| Engagement with the self | At least one self-related construct informed the paper’s research question, theoretical framing, design goal, measure, analysis, or central finding. | “Self” used only technically, such as self-driving or self-supervised learning, or identity, agency, and related constructs mentioned only incidentally. | A paper using self-efficacy as a central outcome was included; a paper mentioning that a system might “increase agency” only as a general benefit, without developing or examining the construct, was excluded. |
Table 1 . Eligibility criteria and examples that mark their boundaries.
Figure 2 . PRISMA Flow Diagram for Screening. PRISMA-ScR flow diagram: 325 records identified, 45 removed before screening, 280 titles screened, 243 abstracts assessed, and 102 papers included.
Figure 3 . The number of self-related papers between 2013 and 2025. Chart of the number of papers in the corpus per year from 2013 to 2025, rising sharply in 2024 and 2025.
| Code | Definition |
| RQ1 — Framings of the Self in HCI | |
| Self as a dynamic process of Transformation (n=46) | The self as something shaped through reflection and ethical self-work, technologies scaffold ongoing change and growth. |
| Self as a Narrative Identity (n=30) | The self as an evolving life story, tools intersect with how people author past, present, and imagined futures. |
| Self as a Social Construct (n=49) | The self as situated in culture and relationships, systems mediate norms, roles, and interpersonal context. |
| Self as a set of Psychological Needs and Capabilities (n=59) | The self as motivation and skill, designs touch autonomy, competence, relatedness, self-efficacy, and self-awareness. |
| Self as a Performance and Act of Disclosure (n=48) | The self as performance to others, interfaces shape impression management and what people reveal. |
Table 2 . Major clusters for RQ1, RQ2, and RQ3. A paper can receive multiple codes.
Figure 4 . Frequency and co-occurrence of the six self framings. Horizontal bars show the total number of papers coded with each framing. Vertical bars show the number of papers with each exact combination indicated by the black dots below; gray dots indicate framings absent from that combination. UpSet plot of six self framings and their co-occurrence across papers.
| Framing | Methods | Scales | Data Types |
| Transformation | Mixed-methods, diary studies, field deployments | Authenticity Scale; PHQ-9/OASIS; Negative Automatic Thoughts (ATQ-N) | Journey maps, diary entries, think-alouds |
| Narrative | Qualitative content analysis, mixed-methods with writing prompts | LIWC; Self-Disclosure Depth (3-level coding); Pennebaker’s Expressive Writing Questionnaire | User-generated narratives, expressive writing, story completions |
| Construct | Participatory co-design, cross-cultural surveys | Social Value Orientation (SVO); Inclusion/Control/Ownership scales; Mayer’s Trust Model | Workshop artifacts, cross-cultural comparisons, co-design outputs |
| Capabilities | Experimental, behavioral quantitative | SDT scales (autonomy, competence, relatedness); Creative Self-Efficacy; NASA Task Load Index (NASA-TLX) | Survey responses, task performance metrics, behavioral logs |
| Presentation | Online experiments, interaction logging, think-aloud | Distress Disclosure Index (DDI); Ownership/Autonomy/Control; Perceived Trustworthiness | Interaction logs, coded disclosure transcripts, text analysis |
| Meaning | Phenomenological, first-person, design explorations | PANAS; Mekler & Hornbæk meaning dimensions; Self-Congruence Scale | Reflective writing, somatic responses, design artifacts |
Table 3. Measurement by self-framing.
Figure 5 . The Inclusion of AI-Self Framework. Four pairs of circles labeled Self and AI, arranged left to right with increasing overlap: Separated, Coordinated, Integrated, and Merged.
| Framing of Self | Typical Research Focus | Indicative System Behaviors | Design Cues & Cautions |
| Transformational | How people become better versions of themselves through AI feedback | Personalized nudges, behavior tracking, habit reinforcement | Use pacing and reflective prompts; avoid overreach into autonomy |
| Narrative | How users make sense of their lives with AI support | Journaling aids, memory recall, autobiographical generation | Preserve authorship boundaries; clarify provenance of synthetic memory |
| Social Context | How identities are shaped in context (culture, history, group) | AI outputs shaped by language, location, or social positioning | Make cultural assumptions explicit; support re-contextualization |
| Psychological | Cognitive/emotional traits and needs AI can measure or model | Emotion classifiers, personality profiling, mood forecasting | Avoid fixed trait assumptions; allow re-calibration over time |
| Presentational | Identity as something performed for others | AI that rewrites bios, curates photos, shapes online persona | Flag editorial changes; allow toggling between authored/self-curated modes |
| Meaning-Making | The self in relation to purpose, ethics, or existential questions | AI that offers advice, spiritual reflection, or existential dialogue | Slow down interactions; use ambiguity to support reflection, not closure |
Table 4 . Operationalizing Self-Framings for AI-Self Research
| Proximity Level | Key Features | User Prompts (diagnostic) | Design Guardrails |
| Separated | Tool-like, limited memory, minimal modeling of self | “I don’t think of it as knowing me.” | Provenance watermarking, minimal personalization |
| Coordinated | Conversational, supports goals, adjusts to user | “It gets how I work, but it doesn’t shape who I am.” | Turn-taking, audit trails, control over learned traits |
| Integrated | Participates in self-understanding or identity formation | “It helps me understand myself.” | Authorship disclosures, opt-outs for generated self-narratives |
| Merged | Blurred boundaries between self and system | “I can’t tell where its voice ends and mine begins.” | Human-in-the-loop review, visibility into system contributions, escalation paths |
Table 5 . Locating AI-Self Proximity in Design