Individual Final Assignment: Paper

Background

Throughout this course, you will work with a team to design an urban AI system based on computer vision. You will also get hands-on experience with machine learning and ethics tools and analyze papers related to the field to understand your design work better. This paper assignment is a way for you to demonstrate what you have learned throughout the course so that we, as instructors, can make a summative assessment.

Learning Objectives

After finishing this task, you will have the ability to present a written argument for a specific set of design choices that aim to (1) develop a concept for an artificial intelligence system in an urban setting and (2) make the proposed design open to contestation to ensure it addresses the needs of citizens.

Instructions

We won’t give step-by-step instructions for writing a paper but will provide tips and suggestions that should help you write a good paper.

  • Write throughout the course. Try to produce raw material in short notes, each encompassing one idea you can adapt later.
  • Create a detailed outline before sitting down to write in full.
  • Use visual material to illustrate your argument, particularly when describing your design.
  • Write in your own words.
  • Have a look at examples of successful papers from previous years’ courses.
  • Use the W-Model (Germano, 2021) for the order you write and revise: beginning, end, middle, end, beginning.

Product

The outcome of this assignment is a paper which adheres to the following requirements:

  • Uses the ACM Master Article Submission Template (single column).
  • Between 3,500 and 5,000 words in length (excluding references, figure/table captions, and appendices).
  • Delivered in PDF format.

The contents of the paper should include:

  • The context you designed for (e.g., the urban issue, the need of the municipality, etc.)
  • A high-level description of the design solution
  • A description of the proposed dataset and model that enables the design solution
  • An analysis of possible reasons why the decisions made by the proposed solution may be contested and by whom (i.e., an analysis of the direct and indirect stakeholders, their values and interests, and how those might be impacted).
  • A description of and rationale for your design’s contestability “features” (a) at design-time and (b) at use-time.
  • A discussion and evaluation of these contestability features, including shortcomings, limitations, etc.

Guidelines for the use of generative AI writing tools:

  • Text generated from a large-scale language model (LLM) such as ChatGPT must be marked where such tools are used for purposes beyond editing the author’s text.
  • We will not use tools to detect LLM-generated text, but we will fail papers where indisputably obvious LLM use is not marked.
  • We recommend that you limit the use of LLMs to “zero-shot translation” tasks, as described by Mittelstadt et al. (2023).

Follow-up

The paper will be graded based on the following criteria:

  • Clarity of the argumentation.
  • Strength of the argumentation.
  • Depths of the exploration of practical and theoretical design issues.
  • Consistency of argumentation with the design concepts.
  • Effective use of supporting literature.