|
Listen to this story
|
Developing and deploying AI responsibly requires tools to help the entire development team understand what they’re doing and show them how their choices affect the end users. Researchers, analysts, and policymakers today are in dire need of minimising, if not avoiding the harm AI model can cause.
Toolkits play a fundamental role in creating systems fairer, more robust and transparent. Here are seven toolkits that can assist in implementing AI ethically.
NASSCOM Responsible AI Resource Kit
In 2022, the National Association of Software and Services Companies (NASSCOM) collaborated with industry leaders including Microsoft, Tata Consultancy Services and IBM Research to introduce the Responsible AI Hub and resource kit. As the name suggests, the objective behind this initiative is to ensure responsible integration of AI technology.
NASSCOM will maintain this evolving reference, continually incorporating the latest research and industry best practices, drawing from several credible sources. This kit equips businesses with the tools and guidance for AI development and deployment while complying with standards ethically. Moreover, the kit offers insights for identifying and mitigating ethical risks that may arise while implementing AI-powered solutions.
AI and data protection risk toolkit
The Information Commissioner’s Office of UK launched an AI and data protection toolkit last year as part of the effort to spread best practices in the use of AI. The toolkit is available as an Excel file on its website to download and edit to help organisations through several stages. During the release, senior policy officer Alister Pearson said it has been developed to help organisations comply with data protection regulations and win public trust in the use of AI.
Ethics in Tech Toolkit
The Markkula Center for Applied Ethics at Santa Clara University developed the project to provide free resources for everyone to integrate ethics into their products and designs. The project includes a comprehensive ethical toolkit for engineering and design practice that consists of seven tools that can be combined with engineering and design workflows. The toolkit can be an attempt to ensure that practitioners responsibly develop technologies and ethics does not become ‘vaporware’.
Playing with AI Fairness: What-if Tool
Abbreviated as WIT, the tool developed by Google makes it easier to examine, evaluate, and debug ML systems easily and accurately. The open-source interactive visual tool enables the understanding of a classification or regression model by enabling users to examine, evaluate, and compare models.
Thanks to its simple user-friendly interface and reduced dependency on coding, this tool caters to a wide range of users. Whether you’re a seasoned developer, researcher, or student, this resource can integrate into your workflow through TensorBoard or as an extension within a Jupyter or Colab notebook.
Ethics and Algorithms Toolkit
The toolkit developed by Joy Bunaguro and his colleagues asks a series of questions that grade the different types of risk in a data-driven initiative. Depending upon the level of risk, the toolkit contains suggestions for mitigations.
Aequitas
Developed in 2018 by the University of Chicago Center for Data Science and Public Policy Aequitas is an open-source bias audit toolkit. The tool is built to be used by a broad set of people from developers to policymakers. The purpose of the tool is to audit machine learning models for discrimination, bias, and make informed and equitable decisions around predictive risk-assessment tools.
Human-AI eXperience (HAX) Toolkit
The tool by Microsoft is best used early in the development process to design human-centric AI systems. The toolkit is made of four parts: HAX Workbook, HAX design patterns, HAX Playbook, HAX Design Library. HAX draws its foundation from a 2019 research paper, effectively turning its theoretical principles into practical tools. In many ways, these guidelines advocate for clarity and precision in UI copy, while being easy to implement plans in case of system failures.
