Guidelines for infusing artificial intelligence to products
Summarizing 20 years of AI design
Many products nowadays are equipped with some functionality enabled by artificial intelligence or AI. We carry AI-infused systems everyday in our mobile devices in the form of activity trackers, mail filters, autocomplete, and social network feed ads. Virtual assistants like Siri, Google Home, and Cortana help us accomplish simple tasks. Recommender systems suggest our next purchase on Amazon, and they direct us to our next favorite series on Netflix.
Nature of AI-infused Products
The common characteristic of AI-infused products is that unlike other systems, they rely on probability and thus involve uncertainty. AI-infused products are inconsistent by nature because they are learning over time. They may also react differently according to lightning conditions, ambient noise, human accents, and other contexts. This unpredictability confuses users, dwindles their trust, and may ultimately lead to abandonment of the product.
Design Guidelines for Interacting with AI
To ensure good design, Amershi and colleagues recommend some guidelines for human-AI interaction. These guidelines leverage more than 20 years of lessons learned in the field. They gathered 168 design recommendations from white papers, reviews, news articles, editorials, and scholarly papers, which they synthesized into the following 18 guidelines.
DG1: Make clear what the system can do. Help the user understand the capabilities of the AI system. For example, when handling user data, be transparent on which data is being tracked and how it is done.
DG2: Make clear how well the system can do what it can do. Help the user understand how often the AI system may make mistakes.
DG3: Time services based on context. The AI system should consider when to act or when to interrupt the user according to their current task and environment.
DG4: Show contextually relevant information. Display information relevant to the user’s current task and environment.
DG5: Match relevant social norms. Deliver the experience in a way that users would expect, given their social and cultural context. For example, voice assistants should be wholesome when talking to families with children.
DG6: Mitigate social biases. Prevent the AI system’s language and behaviors from reinforcing undesirable stereotypes and biases.
DG7: Support efficient invocation. Simplify the method for requesting services to the AI system.
DG8: Support efficient dismissal. Simplify the method for dismissing or ignoring unwanted services.
DG9: Support efficient correction. Simplify the method for editing or recovering from mistakes. For example, after setting a reminder via a voice assistant, users can find the reminder on their phones and they can edit it manually.
DG10: Scope services when in doubt. Engage in disambiguation or gracefully degrade the AI system’s services when uncertain about the user’s goals.
DG11: Make clear why the system did what it did. The AI system should be transparent as to why it behaved as it did.
DG12: Remember recent interactions. Maintain a short term memory and allow the user to make efficient references to it. For example, search engines should understand the context of succeeding queries based on previous queries.
DG13: Learn from user behavior. Personalize the user’s experience by learning from their actions over time.
DG14: Update and adapt cautiously. Limit disruptive changes when updating and adapting the AI system’s behaviors.
DG15: Encourage granular feedback. Enable the user to provide feedback indicating their preferences during regular interaction with the AI system. For example, user’s can mark an email to be important if the AI system missed to classify it as such.
DG16: Convey the consequences of user actions. Immediately update or convey how user actions will impact future behaviors of the AI system.
DG17: Provide global controls. Allow the user to globally customize what the AI system monitors and how it behaves. For example, user’s should be able to switch location sharing on and off.
DG18: Notify users about changes. Inform the user when the AI system adds or updates its capabilities. For example, the AI system can have in-app tutorials for new features.
These guidelines are from the paper Guidelines for Human-AI Interaction published in CHI 2019 by Amershi et al. The method for generating these guidelines and the results of their evaluation are discussed in the paper.
Hi there! Was this article useful?
Writing is my way to remember and share what I learned. 👨🏻💻 It usually takes me hours to read academic papers which I then summarize into 5-minute reads on Medium. Follow me if you are interested in insights from the latest research similar to this article.
Thanks for reading! 🙇🏻♂️
