A mini-guide to initiate stakeholder decision-making about algorithmic impact assessment (AIA) phases
How to use the impact assessment guide
Learn about what the AIA process involves
Complete the questions with your education community to make an informed choice
Reflect upon whether to proceed with this measure at each particular phase
Background
There is currently no education-specific algorithmic impact assessment guide. By documenting the development and impact of algorithmic systems, AIAs can help to mitigate potential harms – especially for vulnerable individuals and communities (Watkins et al., 2021). A report mapping how AIAs are constructed and constitutive components highlights that “there will never be one single AIA process that works for every application or every domain”(Moss et al., 2021: ii). In addition, understanding the following is key to success: who are the stakeholders, the relationship between them, and who is empowered (or disempowered) in terms of whether or not their expertise is represented (ibid).
Overview
Preparation
Learn about the preparation and potential benefits of conducting an impact assessment
Decision point 1
Agree with key stakeholders in your education context whether it is necessary to conduct an IA about an existing, or proposed, technology and whether an IA is necessary (or change of practice)
Planning
Complete the preliminary questions template
Negotiate the planned process, ground rules and expectations with the education community and the edtech vendor
Decision point 2
Decide whether to proceed based on prep and planning phases
Implementation
Initiate the engagement process (e.g. consultations, public log, public forum)
Generate a report which is published openly
Decision point 3
Communicate decisions and recommendations with the education community
Preparation
Decision point 1
Agree with key stakeholders in your education context whether it is necessary to conduct an IA about an existing, or proposed, technology and whether an IA is necessary (or change of practice)
How AIs are initiated
Drivers
Solicited voluntarily by concerned education stakeholders
Mandated by law for the education sector
Triggered by potential harms during system operations at a school or university
Impact assessment benefits
Identifying immediate and short/long-term impacts, harms, and redress procedures related to emerging EdTech systems
Identifying impacts (proxies for harms)
Ameliorating potential harms (allocational and representational)
Recognising lived experiences
Embedding forms of redress and compensation
What is needed to conduct an impact assessment
Essential
Framing: A general framework for conducting IAs in education
Resourcing: administrative and financial resources to conduct an IA for the education sector (e.g. school, government, research, philanthropy, crowdsourcing)
Supporting assessors and time-frame: The individual/team responsible for generating an IA in education throughout the notice, monitoring, and decision-making process
Recommended
Institutional/regulatory body: a government authority to mandate and conduct IAs in education (common, but not universal)
Legislation: to enforce IAs and accountability in education
Planning
Decision point 2
Decide whether to proceed based on prep and planning phases
Preliminary questions
For school stakeholders
What are key areas of concern?
Identify allocative and representational harms.
What resources and opportunities are at stake?
Under what circumstances might someone be denied resources and opportunities as a result of this tool?
Does the tool have the potential to impact particular groups e.g. cultural background, postcode?
For edtech vendors
Are you comfortable with a third party assessment of your tool?
If so, under what conditions?
How to enable inclusive IA methods with diverse education stakeholders
Processes
Public log with ongoing community, researcher, and expert group access to documentation
Consultation via qualitative/quantitative and in-person/online methods and group activities
Public forum to allocate responsibility
Research findings and conclusions published openly
Negotiate the planned process, ground rules and expectations with the education community and the edtech vendor
Implementation
Decision point 3
Communicate decisions and recommendations with the education community
Processes
Public log with ongoing community, researcher, and expert group access to documentation
Consultation via qualitative/quantitative and in-person/online methods and group
Public forum to allocate
Research findings and conclusions published openly
