type
Self-Initiated
role
User Researcher, UI/ UX Designer
timeline
2 months during winter, 2024-25
category
Social Communication / AI Integration
The Brief
The project aimed to redesign the ChatGPT interface based on academic research in human-AI teaming (HAT). It required developing mockups and prototypes with a clear design rationale, demonstrating two use case scenarios and grounding the work in relevant academic literature.
The Outcome
The redesigned interface, named EchoAI, features a dual-channel communication system that enables users to engage in group discussions while simultaneously receiving personalized AI assistance through private channels. The interface includes real-time analytics, contextual awareness features, and adaptive support mechanisms that identify and resolve communication breakdowns.
01
The work
Research
Research into human-AI team dynamics reveals consistent challenges when integrating AI into collaborative environments. Human-AI teams (HATs) typically experience reduced engagement, coordination difficulties, and diminished information exchange compared to human-only teams. Key patterns emerge across studies: transparent AI systems build essential trust, unstructured AI contributions create cognitive overload, imposed AI structures disrupt natural team evolution, and information flow between humans and AI requires deliberate facilitation.
These findings, grounded in established theories of collaborative cognition and communication, inform the dual-channel approach behind EchoAI. The design addresses these challenges by creating separate but interconnected pathways for group conversation and AI assistance. This approach preserves natural team dynamics while providing computational support at critical junctures, transforming theoretical research into practical design that enhances rather than disrupts communication.
Design Rationale
The dual-channel approach differentiates between Human-Computer Interaction (HCI) and Human-Human Interaction (HHI) needs while maintaining connections between them. This architecture responds to research showing that single-channel approaches create tension between AI's structured communication style and human teams' fluid, contextual interactions.
Each design principle addresses specific breakdowns observed in human-AI collaboration, creating an interface that enhances rather than disrupts team dynamics. By optimizing each channel for its primary interaction type while maintaining cross-channel awareness, this approach creates a system that facilitates both proactive information sharing and on-demand assistance while preserving natural conversation dynamics.
Problem Identification
Traditional chatbot interfaces create barriers for effective team collaboration through one-size-fits-all approaches that fail to adapt to team dynamics, limited mechanisms for detecting and resolving miscommunication, cognitive overload from unstructured information, poor contextual awareness across conversation
threads, and reduced engagement when AI is integrated into group settings.
These limitations undermine the establishment of common ground—the shared understanding necessary for coordinated action.
02
Design Process
Ideation
The design process explored several approaches for AI integration in chat interfaces: direct participation in conversations, separated consultation, AI-mediated communication, and parallel interaction channels. Each approach was evaluated against research findings on team coordination, cognitive load, and information exchange patterns.
The dual-channel system emerged as optimal by uniquely enabling simultaneous engagement in both group discussions and private AI interactions. This approach demonstrated superior performance in testing, with higher team coordination scores and satisfaction ratings while preserving natural conversation flow.
Visual Design
The visual design employs a clean aesthetic with a focused color palette of purple, coral, and neutral backgrounds to create clear information hierarchy. Typography combines Montserrat, enhancing approachability while maintaining professionalism.
Consistent structural elements like circular avatars, subtle status indicators, and distinctive purple highlighting for AI content, help users intuitively distinguish between human and AI contributions. White space is deliberately utilized throughout to reduce cognitive load, while the consistent treatment of interactive elements with rounded corners creates a cohesive, accessible interface that requires minimal training to navigate effectively.
Interface components
03
Final Design
Key Features
These six interrelated capabilities operationalize the dual-channel theoretical framework, facilitating both structured information processing and natural social dynamics while maintaining cognitive ergonomics across diverse team interaction patterns.
01
Contextual AI Support
Smoothly integrates AI into group chats through structured summaries and key detail tracking, maintaining conversation flow while supporting shared context development
02
Real-time Assistance
Provides timely suggestions at natural conversation points, guiding decision-making without controlling the interaction or disrupting team dynamics
03
Context Switching
Enables seamless transitions between group discussions and private AI consultations while maintaining relevance across channels and supporting multiple communication paths
04
Information Organization
Actively transforms scattered conversation details into clear, actionable items, reducing cognitive load through structured information presentation
05
Error Detection & Resolution
Identifies confusion in real-time and facilitates resolution through private and group support channels, maintaining social cohesion during miscommunication
06
Resource Access & Adaptive Support
Delivers relevant information through private channels while maintaining group flow, then shifts from problem management to constructive planning once initial issues are resolved
Use case scenario one
Effective Communication
While traditional AI assistants often create an either/or choice between human conversation and AI support, this scenario demonstrates the unique ability to maintain both simultaneously. The sequential process reveals how integrating dual-channel architecture preserves the natural rhythm of human interaction while enhancing it with AI capabilities. This approach directly addresses the fundamental tension in collaborative systems between efficient information processing and authentic social connection, showing how thoughtful interface design can transform theoretical research on human-AI teams into practical, usable systems.
User case scenario two
Resolving Miscommunication
The scenario illustrates critical advancement over traditional interfaces: the ability to detect and resolve communication breakdowns without human initiation. Unlike conventional chatbots that respond only to direct queries, EchoAI proactively identifies emerging coordination issues and adaptively provides assistance based on individual contexts. The five-stage process demonstrates how an interface designed around grounding principles can transform potential team conflicts into opportunities for enhanced collaboration, addressing a key vulnerability in human-AI team integration identified in recent research.
06
Conclusions
Key Takeaways
Design decisions must prioritize team dynamics over AI capabilities – technology should adapt to human communication patterns, not vice versa
Visual language plays a crucial role in building appropriate mental models, with subtle indicators proving more effective than explicit explanations
Separating information density by channel creates cognitive advantages that single-channel approaches cannot achieve
The most effective AI assistance occurs at conversational inflection points rather than through continuous presence
Cross-channel awareness features are essential for maintaining cognitive coherence across different interaction modes
Error detection systems should prioritize private notification before group intervention to preserve social dynamics
Team communication patterns establish fastest when users can easily predict AI behavior through consistent interface design
Reflection
This project fundamentally shifted my understanding of interface design, challenging me to view AI integration not as a technical problem but as a social one. The most difficult aspect was balancing opposing forces: making AI assistance visible enough to be useful but subtle enough to avoid dominating conversation. The visual design process taught me that seemingly minor details like status indicators significantly impact how users conceptualize the AI's role in team dynamics. While the theoretical foundation proved essential, translating abstract concepts like "information pulling" into concrete interface elements demanded creative solutions not found in research alone. Moving forward, I'm interested in exploring how these dual-channel principles might extend beyond text into multimodal environments.
.also see