Continue to Navigate with Timely Feedback
Having the right tools in place facilitating two-way communication is a critical success factor. AI algorithms can be used to continuously collect feedback from customers to improve the gamified insurance experience. Analyzing such feedback data helps insurers identify areas for enhancement or new gamification features that can be introduced for maintaining a high engagement level.
For example, research[1] finds that it is important to recognize the varying levels of individuals and avoid a “one-goal-fit-all” approach. In other words, goals should be personalized, measured from an individualized baseline, and paced such that they can be achieved in more manageable steps, according to progress made along the way. This should help ensure realistic expectations and adherence to the program.
Personalized consultation and regular feedback along with contextually tailored suggestions and well-timed individualized prompts and nudges will inspire consistent behavioral practice and habit formation, an internalization process for promoting autonomy and intrinsic motivation which could be key to long-term commitment. Through more frequent and direct information exchange, insurers can deliver and maintain a more engaging and personalized insurance experience, delivering greater product and customer satisfaction, all the while managing modifiable risks and portfolio value.
The Sapientus Approach
At Sapientus, we are continually assembling a plethora of data from a multitude of sources, including location-based and app-based information as well as a collection of publicly available external auxiliary databases. Through our proprietary machine learning models and computational techniques, we have the ingredients and tools required for supporting our insurer partners in designing and executing a behaviorally informed, active management strategy. For example, one of our behavioral indicators is screen time which has been linked to higher incidence of cardiovascular diseases[2]; we could identify the mitigating factors for managing this behavioral risk through proper incentives.
Some Caveats for Further Consideration
With continued progress and innovation in using emerging data and AI algorithms to better serve customers, there are several factors to be considered as the industry embarks on these various exciting and potentially promising endeavors. Regulatory and ethical considerations must be carefully accounted for prior to actual implementation with real customers. First and foremost, data privacy and security must be protected as per the strictest standards and governance process, ensuring customer consent to data collection and authorized access to the applicable systems, with proper safeguards in place.
Algorithmic bias is another debatable topic. Some argue that AI models are inherently biased, and effort should be focused on managing, as opposed to completely eliminating, such bias.[3] Remediating actions could begin with taking a closer look at the training datasets to review representation of various sensitive attributes. Certain data elements might be highly correlated with a protected trait and introduce potential bias. Generative adversarial network (GAN)[4] and related modeling techniques could be used to expand data diversity and handle intersectionality across data groups, mitigating potential misclassification and unintended discrimination due to data imbalance or other limitations.
Insurers walk a fine line between individualized pricing and excessive price differentiation. Distinguishing between adverse behaviors (e.g., controllable risks such as reckless driving) and unchangeable factors (those covered under the solidarity-based risk pools) is key to avoiding discrimination[5]. Zooming in on factors that customers have a sufficient level of influence over and steering behaviors correspondingly (e.g., through a gamification paradigm) should help insurers reap the benefits of a more granular view on risks while complying with the principles of actuarial fairness, regulatory compliance and societal acceptability.
Human involvement is an indispensable component of the process. Continued monitoring for suspicious patterns, incorporating ongoing feedback with real-world inputs, is key in maintaining the integrity of the algorithms and outputs. Understanding the drivers of model outcomes could help ease regulatory concerns around transparency and explainability, reducing the machine-human gap. Responsible executives and end-users should be made informed and aware of the potential biases and risks. Robust model governance involving appropriate stakeholders from data selection and model development to maintaining model catalogues and audit trails, as well as ongoing risk assessment and management is critical to ensuring equitable practices that serve a diverse population.[6]
A Multidisciplinary Exploration
The concept of gamification is not new to the insurance industry, and we have seen more than a few market examples that inspire further exploration in the subject of digital engagement practices. Success hinges on being able to sustain interest in the “game” with novel elements (e.g., new challenges and mystery rewards) and relevance of the game objectives (i.e., setting manageable goals that matter for the customers) over a long-term period. A test-and-learn approach allows for nimble adjustment and readjustment through repeated experimentation and reinvention. Recruiting experts from different domains, including game design, behavioral science, tech and insurance, would increase the chances for delivering an integrated and seamless experience for customers.
The trifecta effect of emerging data (such as mobile and behavioral data), smart analytics (leveraging AI and machine learning advances) and gamified engagement (through personalized nudges and customizable rewards) could be revolutionary if applied appropriately with proper ethical principles. Insurers can join forces with data and analytics partners in conjunction with UI/ UX experts to curate whole new experiences for their customers – engaging, satisfying, and beneficial for the long-term well-being of all parties.
[1] Mitesh S. Patel, Stacey Chang, and Kevin G. Volpp, “Improving Health Care by Gamifying It” in Harvard Business Review May 2019; https://hbr.org/2019/05/improving-health-care-by-gamifying-it
[2] “Screen-Based Entertainment Time, All-Cause Mortality, and Cardiovascular Events; Population-Based Study With Ongoing Mortality and Hospital Events Follow-Up” in Journal of the American College of Cardiology 2011
[3] “Manage AI Bias Instead of Trying to Eliminate It” MIT Sloan Management Review January 2023
[4] “Generative Adversarial Networks for Actuarial Use” International Actuarial Colloquium May2020
[5] “Regulation of Artificial Intelligence in Insurance: Balancing consumer protection and innovation” The Geneva Association September 2023
[6] “Avoiding Unfair Bias in Insurance Applications of AI Models” SOA Research Institute August 2022
