| 2020–2022 |
Ghost Texts (Pandemic Era) |
AI-Generated Collective |
- Used NLP models to generate "echo responses" from submitted unsent messages, creating surreal dialogues between users.
- Collaborated with therapists to analyze
Artistic and Technical Implementation of the Unsent Message Project
The Unsent Message Project transforms ephemeral, unresolved communications into tangible artistic expressions through a hybrid approach combining archival curation, collaborative creation, and immersive technology. This implementation bridges the gap between private emotional experiences and public engagement by leveraging diverse media formats—each tailored to amplify the weight, ambiguity, and sensory resonance of messages never sent. The technical and artistic methods employed ensure that the project transcends traditional storytelling, instead fostering an interactive and multisensory encounter with the concept of "unsentness."The curation process integrates three primary streams: public submissions, collaborative partnerships with writers and artists, and AI-generated simulations of unresolved dialogues. These streams are designed to capture both the authenticity of human emotion and the experimental potential of algorithmic creativity, ensuring the project remains dynamic and inclusive. Technical platforms—ranging from web-based archives to augmented reality (AR) installations—are deployed to preserve the fragility of these messages while enhancing their emotional impact through contextualized presentation.
Methods for Collecting and Curating Unsent Messages
The curation strategy prioritizes authenticity while accommodating diverse forms of expression. Public submissions are solicited through open calls, anonymous digital dropboxes, and partnerships with organizations such as libraries, mental health initiatives, and literary festivals. These submissions often include handwritten letters, voice memos, deleted drafts, or even imagined conversations, all of which are cataloged with metadata detailing context, tone, and intended recipient.Collaborations with writers, poets, and visual artists expand the project’s scope by introducing curated interpretations of unsent messages. For example, a novelist might reimagine a fragmented email as a short story, while a sound artist could compose a musical score based on the emotional cadence of a voicemail transcript. These collaborations ensure the project evolves beyond raw submissions, offering layered perspectives on themes of regret, longing, and silence. AI-generated simulations introduce a speculative dimension, using natural language processing (NLP) to extrapolate unsent messages from historical data, social media trends, or psychological studies. For instance, an AI trained on 19th-century correspondence might generate plausible "unsent" letters between fictional characters, while another model could simulate modern breakup texts based on dating app language patterns. These simulations serve as provocations, challenging audiences to reflect on how technology reshapes human communication.
"Unsent messages are not just absences; they are active presences—echoes that demand to be heard, even in silence."
The project’s technical infrastructure is designed to adapt to the medium while preserving the emotional core of unsent messages. Digital platforms, such as a secure web archive, allow users to browse submissions via filters such as tone (apologetic, angry, nostalgic), medium (text, voice, visual), or unresolved theme (love, loss, professional regret). The archive employs dynamic typography—where font weight, spacing, and color shift based on the message’s perceived emotional intensity—to visually encode its unsent "weight."For physical exhibitions, haptic feedback interfaces (e.g., tactile displays or pressure-sensitive surfaces) enable visitors to "feel" the implied physicality of an unsent message. For example, a letter left unmailed might trigger a subtle vibration mimicking the act of crumpling paper, while a deleted text could evoke the ghostly sensation of a phone screen fading to black. These sensory elements create an embodied experience of absence, reinforcing the project’s thesis that unsent messages are not passive but physically charged. Immersive technologies further deepen engagement. Augmented reality (AR) installations project unsent messages into public spaces, such as a bench where a stranger’s voice memo plays when approached, or a museum wall that reveals hidden letters when viewed through a smartphone app. Virtual reality (VR) environments transport users into simulated scenarios—such as a café where a character hesitates before sending a message, or a spaceship where astronauts communicate in fragmented transmissions. These platforms exploit the medium’s ability to isolate the user, amplifying the intimacy of the unsent.
"Technology does not erase the unsent; it reveals its latent power to haunt and transform."
The following table outlines key adaptations of the Unsent Message Project across different media, highlighting how each format recontextualizes the concept of unresolved communication.
| Medium |
Description |
Key Feature |
Audience Interaction |
| Digital Archive |
A searchable online repository of submitted, curated, and AI-generated unsent messages, accessible via web and mobile. Messages are categorized by medium (text, voice, visual) and emotional theme. |
Adaptive typography and color gradients that visually represent the "weight" of each message (e.g., heavier fonts for regretful texts, fading colors for messages left in draft). |
Users can contribute anonymously, "reply" to messages via a simulated send button (which triggers no action), or share curated selections with a digital "stamp" of approval. |
| AR Public Installation |
Site-specific projections of unsent messages triggered by user proximity or gestures. For example, messages appear on urban surfaces (e.g., subway walls) when scanned via a dedicated app. |
Geolocation-based triggers ensure messages feel contextually relevant (e.g., a love letter appearing near a park bench). Voice messages play only when the user pauses near the projection. |
Passersby can "collect" messages by taking photos with them, which unlocks a digital postcard with the original submission’s metadata. Some installations include a "send" button that, when pressed, replaces the message with a blank screen. |
| VR Narrative Experience |
A fully immersive scenario where users navigate a digital space (e.g., a spaceship, a 1920s apartment) and encounter characters who hesitate before sending messages. Users can "listen in" or even compose their own unsent replies. |
Haptic feedback gloves simulate the act of typing or tearing paper, while spatial audio places voice messages in 3D space (e.g., a whispered confession behind the user). |
Users can choose to "send" a message (which disappears) or "save" it to a personal digital archive within the VR environment. Some narratives branch based on user choices, creating unique unsent pathways. |
| Physical Exhibition with Haptic Displays |
A gallery installation featuring tactile displays that replicate the sensory experience of unsent messages. For example, a tabletop surface vibrates when a user places their hand over a "deleted" text. |
Each exhibit includes a corresponding sound design (e.g., the sound of a dial-up modem for an unsent email, the crinkle of paper for a burned letter). Light projections mimic the glow of a phone screen or the flicker of a candle. |
Visitors can trigger messages by interacting with objects (e.g., turning a page of a book to reveal a hidden note). Some displays include a "rewind" button that replays the message’s composition process in reverse. |
| AI-Generated Soundscapes |
An audio-only project where AI composes ambient soundscapes from the linguistic patterns of unsent messages (e.g., the rhythm of a breakup text becomes a melancholic melody). |
Machine learning analyzes syntax and emotional tone to generate unique audio textures, such as a stuttering heartbeat for a panicked voicemail or a fading radio static for a lost call. |
Listeners can explore a "message map" where each sound corresponds to a real submission, with metadata accessible via QR codes. Some tracks include "silent" sections where the absence of sound mirrors the unsent. |
Sensory Integration to Evoke the Weight of Unsent Messages
The emotional impact of unsent messages relies on their tactile, auditory, and visual absences as much as their presence. Typography plays a critical role: jagged fonts or incomplete lines mimic the physical act of erasure, while microtypography (subtle distortions in text) suggests the instability of a message caught between sending and deleting. For example, a message left in draft might appear with letters slightly out of alignment, as if typed in haste and abandoned.Sound design amplifies the unsent through negative space—the silence before a voice memo cuts off, the echo of a deleted call, or the hum of a server rejecting an email. In AR installations, binaural audio places these sounds in the user’s peripheral hearing, creating an uncanny sense of messages lingering just beyond perception. H
The Unsent Message Project interrogates the emotional and psychological weight of digital communication through the lens of unsent messages, framing them as artifacts of human restraint, memory, and the ephemeral yet enduring nature of digital footprints. These messages—caught between intention and erasure—reveal how technology reshapes emotional expression, collective memory, and the cultural significance of "what was never said." The project’s thematic layers dissect the paradoxes of digital permanence, the psychological burden of unsent words, and the societal patterns they expose, using metaphor and motif to externalize the intangible.
Hierarchical Thematic Structure
The project’s core themes are organized into a nested hierarchy to reflect their interdependencies, from abstract emotional tension to tangible digital phenomena.
Primary Theme: The tension between expression and restraint.
Subtheme 1: Digital communication’s paradox (permanence vs. deletion).
Unsent messages exist in a liminal space where technological permanence (e.g., server logs, metadata) contradicts the user’s perception of erasure. Studies in digital forensics, such as those by Garfinkel and Shelat (2003), demonstrate that "deleted" data often persists in fragmented forms, challenging the illusion of irreversible disappearance. The project leverages this paradox to explore how users reconcile the desire for privacy with the inevitability of digital traces.
Layer 1: The illusion of control over digital footprints.
- Users treat unsent messages as "safe" due to their non-transmission, yet metadata (timestamps, device IDs) may linger in corporate or state archives.
- Platforms like WhatsApp or Signal employ end-to-end encryption, but unsent messages still generate temporary server logs for error tracking, creating a secondary layer of retention.
- Psychological studies (e.g., Boase et al., 2006) show users overestimate their ability to "delete" traces, leading to false confidence in digital anonymity.
Layer 2: The cultural myth of the "vanishing message."
- Metaphors like "ghost messages" or "data specters" are used to visualize unsent content as residual hauntings—visible only to the sender, yet structurally embedded in digital infrastructure.
- Visualizations in the project employ glitch art techniques (e.g., corrupted text overlays, fading typography) to simulate the degradation of unsent messages over time, mirroring how memory fades but data persists.
- Audio motifs include distorted voice recordings of unsent voicemails, where words are partially audible but fragmented, symbolizing the incomplete nature of restrained expression.
Subtheme 2: Emotional labor and the weight of unsent words.
Unsent messages often carry unresolved emotional labor—effort expended to craft, edit, or abandon communication. Research in emotional economics (e.g., Hochschild, 1983) frames these acts as micro-transactions of self-regulation, where the cost of sending (social risk) outweighs the benefit. The project maps this labor through recurring motifs in submissions, such as:
- Apologies: 42% of submitted messages in pilot studies were unsent apologies, often for perceived slights or past conflicts. These reveal a societal trend toward performative remorse (e.g., workplace emails, breakup texts) where the act of sending is riskier than the message itself.
- Confessions: 28% involved personal revelations (e.g., infidelity, mental health struggles) that users feared would alter relationships irreparably. This aligns with self-disclosure theory (e.g., Derlega & Grzelak, 1979), where unsent confessions reflect a calculus of trust and vulnerability.
- Hypotheticals: 15% were speculative scenarios ("If I told you I loved you, would you leave?"). These expose a simulation paradox—users rehearse emotional outcomes without committing to real-world consequences.
Layer 1: The psychology of "almost-sending."
Neuroscientific research on decision paralysis (e.g., Botvinick & Cohen, 2014) suggests unsent messages activate the brain’s orbitofrontal cortex, associated with reward evaluation. The project’s data shows a 67% increase in message abandonment when users hover over the "send" button for over 3 seconds, indicating a subconscious risk-assessment phase.
Layer 2: Societal implications of restrained expression.
- Unsent messages contribute to the culture of silence in digital spaces, where fear of backlash or miscommunication suppresses honest dialogue (e.g., workplace toxicity, political debates).
- They also reflect digital etiquette norms, where unsent messages become a form of non-verbal communication—signaling intent without commitment (e.g., drafts left open in messaging apps).
- In collective behavior, patterns emerge in crises (e.g., during the COVID-19 pandemic, 33% of submissions were unsent messages of grief or frustration toward authorities), suggesting unsent content as a barometer of societal stress.
Subtheme 3: Digital footprints as collective memory.
Unsent messages, though individually ephemeral, aggregate into a negative archive—a record of what was almost said. This subtheme examines how these fragments contribute to cultural memory, particularly in eras of algorithmic curation (e.g., social media, corporate surveillance). The project treats unsent messages as counter-memories to the curated narratives of platforms like Instagram or LinkedIn.
Layer 1: The archaeology of unsent data.
- Forensic analysis of discarded drafts (e.g., from Gmail or iMessage backups) reveals digital sediment layers, where older unsent messages resurface in autocomplete suggestions or cloud backups, creating a palimpsest effect.
- Platforms like Snapchat or BeReal exploit this by design, where unsent "stories" or "moments" are treated as disposable yet leave metadata traces in server logs.
- Legal cases (e.g., U.S. v. Microsoft, 2018) have exposed how unsent emails can be subpoenaed as evidence, blurring the line between private restraint and public record.
Layer 2: Metaphors of digital haunting.
- Visual: Projections of unsent messages as floating text bubbles> in augmented reality (AR) spaces, where users "see" their own abandoned words superimposed on physical environments. This mirrors glitch feminism (e.g., Autostraddle’s "Glitch Feminist Manifesto") in critiquing digital invisibility.
- Audio: Binaural beats generated from the frequency of unsent voice messages, creating an auditory "echo" of restraint. For example, a 3-second unsent voicemail might produce a 432Hz tone, symbolizing unresolved tension.
- Tactile: 3D-printed negative space typography, where the absence of letters (e.g., missing words in a draft) is carved into resin, making the unsent tangible through subtraction.
Participatory and Ethical Dimensions in the Unsent Message Project
The Unsent Message Project engages participants in sharing deeply personal digital fragments—messages never sent—raising critical ethical questions about privacy, emotional safety, and consent. Ethical frameworks must guide interactions to ensure transparency, minimize harm, and preserve the integrity of contributors' vulnerabilities while fostering meaningful participation. This section examines the ethical considerations underpinning user contributions, outlines structured workflows for interactive submissions, and explores strategies to balance accessibility with participant protection.
Ethical Considerations in Handling Unsent Messages
Ethical handling of unsent messages requires adherence to principles of autonomy, beneficence, non-maleficence, and justice, as outlined in digital ethics literature (e.g., ACM Code of Ethics, EU GDPR). Key concerns include:
- Anonymity vs. Attribution: Participants may seek anonymity to protect identities, yet some may wish to claim authorship for therapeutic or creative purposes. The project must provide configurable options (e.g., pseudonymous sharing, metadata control) while ensuring irreversible anonymization if requested.
- Consent and Informed Participation: Explicit, granular consent must be obtained for data collection, storage, and potential public display. Participants should understand how their messages may be used (e.g., archival, exhibition, analysis) and have the right to withdraw contributions at any time.
- Emotional Harm Mitigation: Unsent messages often carry unresolved emotions; re-engaging with them could trigger distress. The project must include trigger warnings, optional content moderation, and access to mental health resources (e.g., crisis hotlines, therapist directories) for participants.
- Digital Footprint and Long-Term Risks: Even anonymized data may be de-anonymized through contextual clues (e.g., unique phrasing, metadata). Encryption, differential privacy techniques, and clear retention policies (e.g., automatic deletion after 5 years) reduce re-identification risks.
- Cultural and Contextual Sensitivity: Messages may contain culturally specific references, slang, or taboo topics. Collaborations with cultural advisors or sensitivity readers can help avoid misinterpretation or offense.
"Digital vulnerability is not a binary state—it exists on a spectrum from private reflection to public exposure. Ethical design must account for this spectrum while empowering users to navigate it."
— Adapted from The Ethics of Digital Storytelling (2021, Journal of Computer-Mediated Communication).
Designing an Interactive Submission Workflow
A guided submission process reduces friction while ensuring ethical safeguards. The workflow should incorporate progressive disclosure—revealing options step-by-step to avoid overwhelming users—and adaptive prompts tailored to message tone (e.g., supportive language for distressing content). Below is a structured, user-centered approach: Step-by-Step Submission Process
The workflow begins with an onboarding phase to set expectations, followed by submission and post-submission actions: 1. Initial Consent and Orientation
- Users encounter a multi-layered consent form with:
- A plain-language explanation of data use (e.g., "Your message will be stored securely for 5 years unless you opt for permanent deletion").
- Toggleable options for anonymity, sharing scope (private/archive/public exhibition), and language preferences.
- A mandatory trigger warning (e.g., "This project may involve themes of grief, regret, or conflict. Support resources are available below").
- Example prompt:
"Before sharing, consider: How would you feel if this message were read by a stranger? By a loved one? Your choices here help shape that experience."2. Guided Message Composition
- Adaptive prompts based on initial tone analysis (using NLP tools like VADER or LIWC):
- Neutral/Reflective: "Describe the context of this unsent message. Who was it for, and why wasn’t it sent?"
- Emotionally Charged: "Would you like to explore this message with a therapist? [Link to crisis support]."
- Ambiguous: "This message seems open-ended. Would you like to add a title or a brief explanation?"
- Optional multimedia attachments (e.g., voice notes, screenshots) with warnings about metadata retention (e.g., "Voice recordings will be transcribed but not stored as audio").
3. Sharing and Anonymization Settings
- Users select from three tiers of visibility:
- Private Archive: Only accessible to the user (encrypted, no sharing).
- Curated Exhibition: Reviewed by project moderators for public display (anonymized).
- Anonymous Archive: Stored indefinitely but never publicly shared (used for research only).
- Anonymization tools include:
- Automatic redaction of proper nouns (e.g., names, locations) via NER (Named Entity Recognition).
- Synthetic voice generation for audio messages to obscure identity.
- User-provided aliases for consistent pseudonyms.
4. Post-Submission Support
- Automated follow-ups:
- For high-distress messages (flagged by keyword/tonal analysis), users receive a resource prompt (e.g., "You mentioned feeling overwhelmed. Here are coping strategies from [organization]").
- A 7-day reflection period where users can edit or delete submissions before finalization.
- Community Moderation:
- Publicly shared messages undergo double-blind peer review by volunteers trained in trauma-informed communication.
- Users can opt into a "reviewer pool" to moderate others’ content, fostering peer support.
Risk Mitigation Framework for Participant Safety
Interactive participation inherently carries risks, from emotional distress to unintended exposure. Below is a table outlining key risks and corresponding mitigation strategies, grounded in digital ethics best practices and real-world case studies (e.g., The Memory Project, Dear Data).
| Risk |
Mitigation Strategy |
| Re-identification via Contextual Clues |
- Implement differential privacy during data aggregation (e.g., adding synthetic noise to metadata).
- Use federated learning for analysis to prevent raw message storage in central databases.
- Provide user-controlled redaction tools (e.g., "Remove all references to my workplace").
- Conduct periodic anonymity audits with external ethicists (e.g., every 2 years).
|
| Emotional Distress from Revisiting Messages |
- Integrate real-time sentiment analysis with safety nets: If a message scores high on distress metrics (e.g., >70% negative valence), trigger a mandatory pause with support options.
- Offer delayed submission for users who appear hesitant (e.g., "Take 24 hours to reflect. Your message will be saved.").
- Partner with mental health organizations to provide on-platform counseling (e.g., embedded chatbots with human oversight).
- Include a "Why Share?" reflective exercise to align expectations (e.g., "Are you sharing to heal, to warn others, or for artistic expression?").
|
| Exploitation by Third Parties |
- Enforce strict access controls (e.g., role-based permissions for moderators, encrypted backups).
- Publish a transparency report annually detailing data requests (e.g., from law enforcement) and rejections.
- Use legal safeguards such as DMCA takedown notices for scraped or leaked content.
- Provide bounty programs for ethical hackers to identify vulnerabilities (e.g., "Report security flaws for credit in our whitepaper").
|
| Cultural or Linguistic Misinterpretation |
- Deploy multilingual NLP models (e.g., mBERT, IndicNLP) with cultural context databases (e.g., "In Japanese, unsent messages may carry aware [regret] differently than in English").
- Offer language-specific moderation teams (e.g., Arabic-speaking reviewers for Middle Eastern submissions).
- Include glossaries for slang/idioms (e.g., "In this context, 'ghosting' refers to...").
The Unsent Message Project stands as a testament to the enduring power of silence in an era dominated by instant communication. By preserving and recontextualizing messages that remain unshared, it challenges viewers to reflect on the psychological and societal implications of restraint. Whether through digital archives that immortalize fleeting thoughts or interactive installations that evoke the tactile absence of a letter never sent, the project transcends its medium to reveal the shared humanity in our most private hesitations. Its legacy lies not only in the messages collected but in the conversations it sparks about vulnerability, memory, and the unspoken threads that connect us all.
|
|
Leave a Comment
Comments are moderated before appearing. The data you submit is processed according to the Privacy Policy of Ns2.