Echos is a Museum AI Tour App that aims to provide a Deeply Immersive Experience for our audience.

Overview

"Echos" is an HCI project developing an AI museum tour app focused on local museum experiences. Using voice interface design, we explored how conversational AI can create more immersive and personalized museum visits.

Personal Contribution

  1. Initial Research

  2. Concluded user types

  3. Design the Voice User Interface of “Joey” (Personas, Interfaces, User Prompts and Prototype)

  4. Design Documentation

UserType

To personalize the journey, we categorize visitors into Four Distinct User Types and focused on one.

AIBotPersona

Based on the user type, we designed an AI companion that adapts its personality and communication approach to create personalized museum experiences.

UserTests

  1. Core Features & Design Validation

  2. Refining Overall Flow

  3. Finalizing Features & Design Choices

Prototype

Test1: Keyboard Input / Voice Input (A/B)

Key Findings

Key Findings

Key Findings

  • Split preference (50% each) confirmed our multimodal approach was the right decision

  • Voice users valued immersive, conversational flow without breaking focus from artwork

  • Text users prioritized social appropriateness and ease of information retention

Test 2: Input Mode Change Method

Round 1: Initial Discovery

  • Users showed extreme preferences (8/10) for their chosen input method

  • Users won't frequently switch between modes, they'll stick to their preferred method based on personality and context

Test 3: Tap Mode Button

Round 1: Excessive Typing Breaks Immersion (9/10)

  • Users felt constant text input disrupted the museum experience

  • Long artwork names were hard to remember and type out ("Tablet with Instructions for the Deceased in the Underworld")

  • Users wanted clickable options instead of typing everything

Round 2: Horizontal Card Scroll - Mental Model Mismatch (9/10)

  • Option cards presented horizontally didn't align with chat app conventions

  • Users expected vertical scrolling like typical messaging interfaces

  • Horizontal interaction felt foreign in a chat context

Round 3: Card Format Confusion (7/10)

  • While users could understand it's a card layout, they felt compelled to pick one due to task framing

  • Cards = Information display in users' mental model

  • Buttons = Action selection in users' mental model

Round 3: Simple Binary Choices (6/10)

  • For yes/no questions, users found typing unnecessary friction

  • Expected direct button selection for simple responses

  • "It's easier to choose instead of type"

Round 3: Interaction Design Refinement

  • Drag interaction didn't match users' mental model (7/10)

  • Show the mode switcher only during active input to maintain clean UI while ensuring accessibility when needed