Ethics & Bias in AI Module Assignment

Assignment introduction

In this piece of work you will produce an Ethical Impact Assessment of a piece of technology that  utilises AI and develop a set of recommendations in response to your findings. You will use Value Sensitive Design to inform your approach.

Assignment format and submission

You should submit your Ethical impact assessment and recommendations report via Turnitin. The

word limit is 1500 words, excluding tables and bibliography. The deadline for your submissions is 2pm on Tuesday 19 March 2024.

Please note that all submissions will be subject to plagiarism and collusion checks. If you use any generative AI tools you must clearly reference these.

Marking overview

Assignment descriptions and tasks

While working for a start-up, you are leading a machine learning project that involves designing an algorithm to perform an emotion recognition task on images of human faces. You can specify a more constrained application domain or use context. You have to produce an ethical impact assessment and a set of recommendations for the project before work commences. Specifically, you are required to analyse the task of emotion recognition using Value Sensitive Design and provide a set of recommendations that makes use of existing guidelines and toolkits.

Your report should include three sections:

-       Introduction

-       Ethical impact assessment informed by Value Sensitive Design

-       Recommendations

Details on what should be included in each section are detailed below.

1a) Introduction (15 MARKS)

In the introduction you will introduce the task, convey to colleagues in the company why ethical impact assessments are useful and summarise the Value Sensitive Design approach. You should anticipate that the system you produce could be deployed by various different organisations or be made available to others via open sourcing.

In this section you should:

•     Provide a description of what Value Sensitive Design is, providing arguments for why it is appropriate or beneficial for this task.

•     Provide a description of the machine learning task you are evaluating, the data sources that will be used to develop the system, the types of predictions or outputs it will produce.

If you are focusing on a particular use-context, clearly introduce this.

•     Provide details of potential use cases of the system and any immediately apparent ethical issues that should be considered in advance of development.

1b): Ethical impact assessment informed by Value Sensitive Design (35 MARKS)

For the ethical impact assessment you need to identify the key stakeholders and values relevant to the project.

You can present the following information in a table:

•     Identify key stakeholders and differentiate between direct and indirect stakeholders

•     Detail which values are relevant to different stakeholder groups in this project

•    Assess and describe key risks and potential harms for each stakeholder group

Example table:

Additionally, you should include a short-written analysis of the key value conflicts.

•     Explain how you identified / selected the human values and provide any additional context on why they are relevant to the particular stakeholders.

•     Provide a summary of the value conflict analysis. Identify and describe any potential value

conflicts and provide an argument for how these may be resolved, or which group’s interests should be prioritised.

•    Articulate any additional considerations or actions. Outline any additional empirical or

technical investigations you would need to carry out to complete your assessment. Can you identify any potential areas that might pose an issue (you can draw on similar or related examples from the literature or case studies where appropriate)?