Skip to main content Link Search Menu Expand Document (external link)
AI & Society

Workshop: Map Your System & Add a Use-Time Contestability Loop

Background

Designing urban AI systems can be an overwhelming task. There are many parts and actions to consider, and it is hard to keep them all in your head. When we want to be responsible about the design of these systems, we do need to consider them as a whole. Furthermore, when we want to ensure contestability of the system at use-time, we need to be able to “see” the system so we can figure out where best to make the intervention happen. In this exercise we take a technique from the field of service design called “service blueprint”, and use it to map out our system, as well as its contestability loop at use time.   

Learning Objectives

After this exercise you will be able to design a contestability loop at use-time that can be added to a new or existing urban AI system.

Instructions

  1. Create a service blueprint of your system:
    1. Review your list of stakeholders, and make sure you have identified the core (direct) stakeholder. You can think of this more or less as the traditional user, even though they may not have any 1:1 interaction with your system. For example, the citizen who wants to just get rid of their trash is a main stakeholder in a stray trash detection system, but they don’t use it directly.
    2. If you haven’t yet, establish what the ultimate goal of the core stakeholder is with regards to your system  – e.g. “keep home waste-free with minimum hassle.”
    3. Grab the empty canvas attached to this page and add it to your Miro board
      • Notice the three horizontal lines across your map, from top to bottom these are: (1) the line of interaction – where user and system interact; (2) the line of visibility – beyond which the user can no longer “see” into the system; and (3) the line of internal responsibility – where the system hands over an action to an “external” or auxiliary system. 
      • Also note the swim lanes: (1) user actions; (2) frontline worker actions; (3) back-office worker actions; (4) AI system actions; (5) third-party actions 
    4. Create a timeline of the lower-level actions your system’s user (and/or individual citizen) moves through as they attempt to achieve their overall goal – e.g. “collect trash in waste bin” or “place trash on curb.”
    5. For each step of this timeline, consider what your system must do to support that goal – list each action roughly underneath the corresponding user action – note that these actions are not necessarily performed by a machine but can also be performed by a human.
    6. Place the actions your system takes below the appropriate lines
    7. Connect actions that are related with lines
    8. Next to the relevant actions, make a note of any physical/material touchpoints/objects/resources that are involved with the specific action – e.g. “scan car”
  2. Add a contestability loop to your blueprint:
    1. Add an action in the citizen/user’s swimlane for “understand system decision” and another one for “request human intervention”
    2. Place the actions where you think they are most likely to happen on the timeline
    3. Specify the actions the system needs to take in support of and in response to these user actions, and place them in their appropriate spots on the blueprint
    4. Once again, make a note of physical/material resources that are involved with these system actions  

Product

Upon completion of this activity you will have created a blueprint that describes how your envisioned system will work, on a conceptual level. Your flowchart will also include a “contestability loop” at use-time – a means for people affected by system decisions to exercise their right to human intervention.

Follow-up

We will discuss your findings from this workshop in the plenary discussion later in the day.

Downloads