Explainable AI Can Assist Internet of Behavior Efforts

The application of explainable artificial intelligence to Internet of Behavior techniques may help provide a more trusted and understandable framework in changing human behaviors, researchers say. This combination of Internet of Things devices, artificial intelligence, data analytics and behavioral science can also achieve user and business benefits, according to a study.

U.K.-based researcher Haya Elayan delved into the concept of Internet of Behaviors (IoB) and its integration with explainable artificial intelligence (XAI) algorithms to observe the impact of changing Internet of Things-related behavior. IoB is the confluence of Internet of Things (IoT) devices combined with advanced sensors, computer vision, facial recognition, other biometric indicators or location tracking, which, when added to insights from human psychology and data analysis, can influence, prompt or change human behavior—a powerful method.

“Nowadays, the use of the IoT, cloud computing and artificial intelligence (AI) have made it easier to track and change the behavior of users through changing IoT behavior,” she states. Elayan presented her findings in September 2021 in the IEEE study, Internet of Behavior (IoB) and Explainable AI Systems for Influencing IoT Behavior, along with Moayad Aloqaily and Mohsen Guizani, professors from Mohamed bin Zayed‌ University of Artificial Intelligence in Abu Dhabi.

Elayan, who is now pursuing a Master of Data Science degree at Leeds University in the United Kingdom and is also a research scientist for xAnalytics Inc., headquartered in Ottawa, Canada, created an XAI platform and leveraged the use of the IoB to apply to a household electricity consumption end-use case. The research was aimed at changing consumer behavior into more eco-friendly consumption patterns, thereby reducing power use, energy waste and costs. The application of such a platform showed initial cost and power savings by users, she reports.

“A system based on IoB and XAI has been proposed in a use-case scenario of electrical power consumption that aims to influence user-consuming behavior to reduce power consumption and cost,” she says in the study. “The scenario results showed a decrease of 522.2 kilowatts of active power when compared to original consumption over a 200-hour period. It also showed a total power cost saving of 95.04 Euro for the same period.”

Elayan asserts that during natural disasters or crises, in particular—such as a global pandemic—IoB and XAI can be used in meaningful ways to help improve or save lives. “Many behavioral aspects have been altered by the COVID-19 pandemic, such as customer interaction with brands, employee work procedures and business engagement with consumers,” the researcher says. “All of these and other examples have an economic, technological and physiological effect. Therefore, tracking people’s behavior becomes crucial to influence it in adverse situations. For example, using machine learning for mask recognition tasks is one way to get individuals to respect regulations and monitor negligence.”

The data scientist experienced an IoB-related environment firsthand when she was home in Jordan during the pandemic; the experience led to the genesis of her research. “I was using apps for tracking my health status and was quarantining based on data from the application,” Elayan explains. “If someone in our office got infected, we would get notification to quarantine because we were in the same place with someone that was infected. That experience made me want to understand what the tracking [aspect] was and how was it influencing me. This is how the process started for me, generating my interest.”

It is the addition of XAI that makes the difference, she says, as it provides a layer of understanding to users, building their trust during the prompted behavioral change process. “If we are using AI to analyze users’ behavior, you can easily use XAI to let the people understand what this AI model does and why, and just giving a better perception of the system,” she explains. “Therefore, the process of tracking, analyzing and influencing the behavior will become much simpler because you are developing a trustworthy platform, with the user understanding what is happening, how it is happening and what is happening.”

The explainable approach also helps to overcome a common challenge in the use of IoB referred to as the ostrich effect. “There is a challenge for Internet of Behaviors called the ostrich effect, with people essentially being afraid,” she says. “And you may encounter resistance when you are trying to influence and change people’s behavior because it’s a sensitive area. You are dealing with sensitive data, and you are dealing with their behaviors. In order not to encounter this, or to manage this resistance and other psychological factors related to comfort and stress, XAI helps us provide the user with the required understanding and trust for the system that uses an AI model.” 

Elayan and the research team relied on IBM’s AI Explainability 360 to help provide explanations of the AI model. The study also had to make sure cybersecurity elements protected the moving data elements, including anything sent back to the users, which is a key component of IoB being able to influence their behavior. “Because we are transmitting the users’ data, we are analyzing it, we are saving it, we’re sending the results back to the user. It was difficult to determine which security scheme to use,” Elayan acknowledges.

However, selecting the appropriate artificial intelligence algorithm was not straightforward, the researcher emphasizes. “[Another challenge] was using or choosing the right AI technique to implement in this system, because the data structure is kind of different, the system and how it connected with our architecture [was unique],” Elayan suggests. “And you are using different technologies, you are using Internet of Things, you are using smart meters, and you have end users. So, what is the correct AI technique to use in order to speed up the process, keep the data safe and apply security schemes and technologies at different levels and processes along the system?”

The lack of IoT device hardening and cybersecurity was another complication. “From an IoT perspective the security schemes are limited,” she notes. “Sometimes you need to develop your own security schemes in order to preserve the data on the IoT devices.”  

Another issue was the lack of regulations and standards related to the ethical use of the data, Elayan continues. “One of Internet of Behavior’s major challenges is how the companies apply users’ data and how they analyze it. Users should be aware of that because companies most of the time are trying to achieve their own benefits and gain more money, and maybe they misuse or abuse users’ data [when] changing behavior. To gain the benefit, they may be hurting the end customer just to gain more money.”

IoB can be a powerful tool, given that it focuses on human nature. The behavioral aspect of humans, and not necessarily other characteristics—such as cognition, emotion, personality and communication—is responsible for the tendency to act, and when combined with digital networks and devices is a powerful factor. First discussing IoB in 2012, University of Helsinki Department of Psychology Professor Göte Nyman reportedly explained that if human behavior was assigned to devices with specific addresses, there would be an opportunity to benefit from the knowledge gained by analyzing the history of patterns in many businesses, societal, health, political and many other fields. More than 10 years later, the IoT sensor commercial market is expected to reach $22.48 billion by 2023, with a predicted 29.3 billion connected devices available, according to a CISCO-produced study the researcher cites.

Elayan is interested in applying IoB to the medical industry, which would require additional standards and cyber measures, she emphasizes. IoB, like any digitally based effort, when paired with patient data, could be prone to attacks.  

“Behavioral data is a sensitive and personal data type, and its collection, storage or analysis must be accompanied by transparency and ethical use,” she notes. “The user has the right to be aware of this process as well as to know that their privacy is preserved and protected from misuse. But focusing on the behavior will allow us to know how to influence and treat the person,” Elayan says.

“I am keen on applying this technology to health care because I know that we have promising results in saving electricity and I’m sure that applying it in healthcare will add great value.”