In a rather imperceptible way in the mainstream media, social science models and technology’s ability to analyze and predict human behavior are converging.
This area – known as convergent technologies – is developing at a rapid pace. So much so that, according to Fujitsu’s forecast, in 2024 we will witness unprecedented insights and the field of social modeling supported by artificial intelligence will be created.
In 2024, for example, with the help of AI-supported digital twins, they will try to deal with social challenges such as climate neutrality and the circular economy. The ability to use predictive behavioral models based on observed data from real people instead of potentially biasing theoretical models will completely transform this field, allowing extremely precise customization.
The idea of convergence between the two fields is not a new phenomenon. In 2019, the Cleverism an article was published in the magazine, according to which society and culture create science, and science in turn affects society. A 2021 academic thesis also discussed the intersection of artificial intelligence and social theory, citing as a difficulty the fact that data is not uniformly available for different areas and theories of social life. However, this problem is no longer an obstacle. The computer image recognition capabilities of artificial intelligence have now advanced to such a level that they can recognize and even predict human behavior.
THE Fujitsu Actlyzer artificial intelligence, for example, understands, predicts and evaluates things in a human-like way. Acting as the “eyes” of the machine, the technology records the elements of human behavior directly from the camera images. Based on this, the AI understands the situation and can predict what will happen next. Keeping privacy in mind, we are also developing technology that can analyze human behavior without video recordings. We are also working on more accurate human modeling technology using the results of the humanities and social sciences, such as behavioral science and psychology.
Access to large amounts of behavioral data and advanced analytics provide ever-deeper insights into collective dynamics, network operations, and complex social patterns. And the new models born in this way precisely reveal the mechanisms behind the manifestations of the herd spirit, for example in relation to the operation of financial markets, the spread of misinformation and the acceptance of social norms. If these models are displayed in digital twins, the intervention “What if?” for the purpose of type analysis, we can obtain information that underpins policy decision-making.
Modeling and computational social science will simulate societies in unprecedented detail, enabling high-level testing of theories and policies. Analyzing the dynamics of a city and the needs of its inhabitants, for example, can help design housing, mobility and recreation strategies adapted to local conditions. And the new pedagogical models enable personalized, AI-supplemented education that more effectively supports different learning styles.
The age of digital trials
We’re not there yet, but we’re not too far away. We are starting to see that these possibilities can actually be realized. In April, Fujitsu presented a new technology called “digital rehearsal” that supports the development of policies and business planning. The first demo carried out in cooperation with Beryl, which provides shared mobility services, provides real simulations of the effects of different transport policy strategies. It uses a digital twin to replicate how people use e-scooter services on the Isle of Wight in the UK.
The new technology is the economic “prospect theory” that models human behavior (Prospect Theory) and by combining artificial intelligence, it draws conclusions about the real behavior of people. It reproduces biases due to human biases that tend to overestimate loss and underestimate potential gain, as well as situational factors that influence behavior, such as the weather. A digital trial on the Isle of Wight is pre-testing the impact of people switching from cars to electric scooters. Fujitsu also estimates the impact that replacing cars with e-scooters would have on CO2 emissions, and to what extent various measures (e.g. service fee discounts for users who return the scooter to a specified point) would influence people’s choice of means of transport. The ultimate aim is to provide business benefits for Beryl, reduce the adverse environmental and social impacts of car use, support transport policy decisions and make a positive contribution to the wider economy of the Isle of Wight.
Better behavior prediction and social modeling
The above is just one example. We can expect rapid development in other areas of application of behavior prediction: improving well-being and realizing a safe and sustainable society, but there are also promising developments in crime prevention, for example.
Experiments are currently underway in Japan to help prevent fraud by analyzing the physiological data of potential victims of telephone scams. The technology estimates whether someone is being deceived based on physiological data indicating changes in feelings related to anxiety (e.g. breathing, heart rate).
Of course, this approach has its risks. Privacy must be guaranteed and distortion caused by bias must be eliminated. However, the current practice of social modeling is perhaps even more at risk of erroneous conclusions due to small data samples and the potential for researcher bias. AI-based social modeling can make predictions about human and social behavior more accurate and thereby improve and rationalize the operation of social services.