MLUX Case Study: Microsoft Research’s Human-AI eXperience (HAX) Toolkit
The 2021 edition of the Machine Learning and UX Meetup (“MLUX”) ended on a great note with a talk from Dr. Saleema Amershi, Senior Principal Research Manager at Microsoft Research, and Lead of the Human-AI eXperiences (“HAX”) team. In this talk, Dr. Amershi covered the HAX Toolkit, which supports building fluid and responsible AI experiences. She also walked through a case study of how a team within Microsoft used different components of the HAX Toolkit to design a human-centered, AI-based feature.
So, what is the HAX toolkit?
The goal of HAX Toolkit is to support day-to-day practices of interdisciplinary teams working at the intersection of HCI and AI. The first tool is ‘Guidelines for Human-AI Interaction’ — a set of 18 generally applicable guidelines spread across different stages of interaction (initially, during the interaction, when the AI is wrong, and how AI should evolve over time).
These guidelines are both published in ACM CHI 2019 and available as part of the HAX toolkit.
To better integrate guidelines into development workflows, other tools such as the Workbook, Design Patterns & Library, and Playbook were developed.
The Microsoft Family Safety App, designed to promote digital safety and wellbeing for families, uses an AI-powered feature. This is the ‘Flagged’ search’ feature which proactively flags or highlights concerning content in web searches. The HAX Workbook helps in the ‘planning stage’ by acting as a discussion guide for the interdisciplinary team. Using the workbook, teams can select relevant human-AI guidelines, map it to implementation requirements, prioritize and track them. Thus, the workbook facilitates early and upfront discussion of best practices.
In the ‘design phase’, the HAX Design Library serves as a useful resource. It is a collection of design patterns and examples that demonstrate how the guidelines and patterns have been realized. You can filter by guideline and product area (ex: advertising, e-commerce, email etc) to narrow your search for examples.
In the ‘prototype stage’, it is important to test ideas and gather feedback. When working with AI, it can be hard to anticipate if and when an AI system might fail and how we can design to mitigate it. The HAX Playbook was developed to helps UX professionals in this stage by allowing for interactive exploration of common AI failure scenarios. By providing failure scenarios for testing, designers can account for them early in the process.
Hear it in Saleema’s own words, watch to the entire talk, including a detailed case study on the Microsfot Family Safety App and Q&A, on our MLUX youtube channel:
So, why is MLUX excited about the HAX Toolkit?
Yes, we love a good toolkit that inspires data scientists, AI researchers, ML engs, UX’ers to collaborate and develop a shared vocabulary — but the real reason we love the HAX Toolkit, is that anyone can also contribute to it since all of these resources are open-source! This means that YOU can submit examples, patterns, scenarios from your team to different components of the HAX Toolkit Design Library.
Reach out for more information, providing feedback or submitting examples and patterns here: https://www.microsoft.com/en-us/haxtoolkit/contact/
The talk ended on a positive note about the Responsible AI movement we have been seeing in industry circles and the importance of interdisciplinary work to ensure human-centered AI experiences.
This is the last talk for 2021 but stay tuned for events in 2022! 👀
About the Machine Learning and User Experience (“MLUX”) Meetup
We’re excited about creating a future of human-centered smart products, and we believe the first step to doing this is to connect UX and Data Science/Machine Learning folks to get together and learn from each other at regular meetups, tech talks, panels, and events (held remotely).
Interested to learn more? Join our meetup, be the first in the know about our events by joining our mailing list, watch past events on our youtube channel, and follow us on twitter (@mluxmeetup) and Linkedin.
