Here are the main points from the Tent Talks session with Amy Bucher titled, “Personalizing Behavior Change: The Intersection of AI, Ethics, and Team Collaboration:”

Session Introduction:

• Amy Bucher discusses personalization in behavior change.
• Highlights the importance of tailoring communication to each individual's context and needs.
• Explores the use of reinforcement learning in personalization.
• Distinguishes between broad, AI-driven approaches and audience-specific design.

Defining Personalization:

• Personalization involves communicating with individuals as if you were talking to them one on one.
• It requires understanding an individual's context, preferences, and history.
• Goes beyond segmentation and adapts to changes over time and across contexts.
• Acknowledges that personalization should account for situational variations in behavior.

The Role of Reinforcement Learning:

• Lirio employs behavioral reinforcement learning, a subset of AI.
• Reinforcement learning involves designating outcomes to reward the algorithm for achieving.
• In healthcare, the primary goal is to reward behaviors like completing a mammogram.
• Reinforcement learning is advantageous for its ability to learn individual characteristics over time.
• It can help overcome biases present in manually designed interventions.
• Amy emphasizes the importance of having data-attached behavioral outcomes for effective reinforcement learning.

Benefits and Challenges of AI in Personalization:

• AI, specifically reinforcement learning, allows for broad, data-driven design.
• It enables the inclusion of various elements in interventions to cater to different subsets of the population.
• It's especially valuable in healthcare for tracking outcomes like appointment attendance.
• Challenges include detecting behaviors without data and maintaining privacy.
• Amy emphasizes the need for finesse in designing for more elusive behaviors like exercise.
• She also discusses the benefits of AI in overcoming biases and the validation of AI-driven messaging through behavioral science.

Dealing with Limited Data:

• Amy discusses approaches to personalization with limited data.
• Mention of "transfer learning," where knowledge about one person is used to inform personalization for another with similarities.
• Emphasizes the importance of minimizing data hoarding and maintaining privacy.
• Talks about using demographic data and behavior-influencing demographics for "warm starts" in AI personalization.
• Stresses that interactions with individuals quickly replace initial data, making personalization more accurate over time.

Ethical Considerations:

• Addresses the ethical question of why an AI should recommend actions to individuals.
• Draws on motivational psychology, discussing "volitional non-adherence" – allowing individuals not ready for action to make their own choices.
• Highlights the risk of resentment and short-term gains with forceful recommendations.
• Discusses the need to respect individual autonomy in healthcare decisions.
• Explains the importance of maintaining trust with individuals and not pushing them into actions they're not prepared for.

Balancing Personalization and Intrusiveness:

• Discusses the need to obtain permission for data usage and communication methods.
• Emphasizes data privacy rules and regulations and securing data with certifications.
• Highlights the importance of "white-labeling" interventions to maintain trust with clients and end-users.
• Considers the "creepiness" factor and how personalization varies depending on communication channels.
• Discusses using first names in emails as an effective personalization strategy.
• Notes that personalization effectiveness may vary across different communication modalities.

Navigating Interdisciplinary Teams:

• Amy leads the behavioral science team at Lirio, working closely with AI and platform teams.
• Emphasizes the importance of asking questions and seeking clarity, even if it means sounding "stupid" in interdisciplinary discussions.
• Stresses the value of cross-functional meetings and sharing ideas early in the design process.
• Reflects on her experience with the platform team when designing for colonoscopies and the importance of aligning behavioral requirements with technology capabilities.
• Highlights the value of collaboration among leaders of different teams.

Types of AI for Different Design Needs:

• Discusses reinforcement learning as used by Lirio for behavioral outcomes.
• Touches on large language models like Chat GPT and Google's BERT as tools for generating content.
• Explains experiments with generative AI for content generation.
• Mentions recommender systems, such as those used in e-commerce, and their potential in healthcare for lifestyle management.
• Envisions using natural language processing (NLP) for bidirectional messaging and understanding user intent.
• Stresses the importance of managing risks when using AI for content generation, particularly in healthcare.

Notable Quotes:

"I think we really have to be okay with them still not being ready to do that. We can provide them the reasons we can have that gentle conversation with them. But ultimately, if they don't want to take action, I think that's something we just have to learn from and move on."

"I think it's so helpful to have people at the top of the teams who are willing to work with each other."

"The risk of delivering somebody the wrong message is so high. Whether they took action on something that's not really good for them, or they missed a recommendation that would have been incredibly helpful or even life-saving."

"We really want to understand the role that [generative AI] might play, but we have some concerns about it."

"One of our Lirio maxims that people say all the time, share before you're ready."

"You really have to be willing to sound stupid in some of these conversations."

Chicago Camps hosts irregularly scheduled Tent Talks with people from all across the User Experience Design community, and beyond. Who really likes limits, anyway--If it's a cool idea, we'd love to hear about it and share it!

What is a Tent Talk? That's a great question, we'd love to tell you.

Tent Talks are short-form in nature, generally lasting from 10-20 minutes (ish) in a recorded format--we like to think of them as "S'mores-sized content" because that's pretty on-brand. Tent Talks can be a presentation on a topic, a live Q&A session about the work we do, or the work around the work we do, or really just about anything--we don't want to limit ourselves, or you.

You should send along an idea or topic of your own so we can learn from you, as well! You don't have to be a published author or a professional speaker on a circuit to be good at your job, so please, put yourself forward, and let's have some fun, talk, and share your experience with others!