the-midnight-paper

Authors: Shiri Azenkot, Catherine Feng, and Maya Cakmak

ACM Human Robot Interaction

What is the high-level research question

How a building service robot could interact with and guide a blind person through a building in an effective and socially acceptable way

Specific Research Question and Hypothesis

RQ1: How should a BSR approach a blind person?
RQ2: What kind of feedback should a BSR provide while guiding a blind person through a building?
RQ3: How should a BSR interact with bystanders when guiding a blind person?

Key Findings

RQ1: How should a BSR approach a blind person?
- User should be able to (1) find out whether there was a robot in the building, and (2) summon the robot to her current location.
- A robot that singled out a person with a mobility aid may be perceived as paternalistic or simply annoying.
- Since many blind people are smartphone users, they should be able to make contact with the robot through a smartphone application.
RQ2: What kind of feedback should a BSR provide while guiding a blind person through a building?
- Robot should be “friendly” and act “like a receptionist”.
- The robot should describe the different kinds of assistance it can offer, allowing the user to customize her experience.
- Information Kiosk: In this mode of assistance, the robot would only provide information and routing instructions, without accompanying the user to a destination.
- Escort: In this mode, the robot would guide the user to a destination by walking next to or slightly ahead of the user, without direct contact.
- Sighted guide: This mode emulates a human sighted guide. The robot guides the user to her destination by walking next to and slightly in front of the user.
- While walking with a user in Sighted Guide Mode, the robot should adjust its pace according to the user.
- The robot should use landmarks that are accessible to blind people, such as audible or tactile cues (a water fountain, coffee bar, or the edge of a carpet).
- The robot should verbally direct the user when walking in Sighted Guide and Escort mode.
RQ3: How should a BSR interact with bystanders when guiding a blind person?
- robot should behave differently when encountering people than inanimate objects
- the robot should behave in a socially appropriate manner, speaking softly but audibly and politely to the user and the people ahead.
- Optionally, robot should let blind people speak for themselves by giving them detailed descriptions of the surroundings in all assistance modes

Data Collection Methodologies

Robot Used: PR2 Authors used participatory design where all stakeholders were involved in the design of the system. They designed the system collaboratively with designers and target users. Design team consisted of three designers and five non-designers with varying visual abilities. Two of the designers had vision impairments. Blind people have a wide range of vision abilities and mobility skills and therefore there is no single mode of assistance to fit needs of all visually impared users.

The design was created over multiple sessions:

A 30-minute one-on-one interview conducted by a designer with each non-designer. Goals were to
- describe the structure of the project,
- describe the robot’s assumed capabilities and our design space, and
- probe the non- designer for ideas.
A group brainstorm and discussion session with all non-designers and two designers
- The goal of the group session was to develop a conceptual storyboard that described the interactions between the robot guide and the user, from the time they met to the time they separated.
- How will the robot and the blind person connect?
- How should the robot guide the person?
- What kind of feedback should the robot provide while it’s guiding the person?
- What should the robot do when reaching a destination?
An individual design sessions with a robot, performed by a designer and each of the non-designers. Prototyping robot’s navigation was done with Wizard-of-Oz, where a designer controlled the robot with a joystick and spoke for it.
- The goal of the third session was to develop low-level specifications for how the robot should behave when guiding the user.
- How should the robot move when guiding the user through a doorway?
- What should the robot do when approaching and passing a turn or an obstacle in the path?
- When exactly should the robot give feedback when approaching a turn or obstacle in a hallway?

Interviews were conducted remotely and all other sessions were conducted in a university lab containing the robot.

Data Analysis Methodologies

Authors audio-recorded all sessions and took notes, then transcribed the recordings, coded the transcripts, and organized the codes by common themes.

Did the paper draw convincing conclusions using the methodology

yes?