TEXT GENERATION
AI- FINE TUNING
STYLE MIMICRY
PATTERN REGOGNITION
ROLE
Data Engineering
Data Visualisation
Data Engineering
Data Visualisation
TIMELINE
2 Weeks
2 Weeks
SKILLS
Data Training
Data Training
Can AI help us write something truly meaningful?
OVERVIEW
Over the course of this project, I designed and built a human–AI co-creation system that generates love letters using a fine-tuned GPT-2 model trained on curated datasets and literary styles.
PROBLEM
As generative AI becomes more present in personal communication, this project explores how AI can support emotional expression without taking away human intent.
As generative AI becomes more present in personal communication, this project explores how AI can support emotional expression without taking away human intent.
DESIGN FOCUS
Human AI Collaboration
Designing interactions where users guide and shape AI-generated content rather than passively receiving it.
Designing interactions where users guide and shape AI-generated content rather than passively receiving it.
Authorship & Control
Exploring how editing, iteration, and tone selection influence a user’s sense of ownership.
Exploring how editing, iteration, and tone selection influence a user’s sense of ownership.
Emotional Trust
Understanding how AI-generated writing can feel authentic, personal, and meaningful.
Understanding how AI-generated writing can feel authentic, personal, and meaningful.
GOALS
Increase user control over tone, style, and output, Design a system that enables human–AI
co-creation instead of full automation
co-creation instead of full automation
Build trust in AI-generated writing through interaction and transparency
EXPLORATION
I began by exploring different ways AI could participate in emotional writing. Early ideas ranged from fully automated letter generation to guided writing assistants and collaborative editing systems.Through sketching and mapping user journeys, I explored questions like:
Should the AI generate complete letters or assist line-by-line?
How much control should the user have?
How can the system feel personal rather than mechanical?
I moved away from fully automated approaches and focused on designing a collaborative system where users remain central to the writing process.
