The CREF-Sony Joint Initiative (JICS) is unique in that it unites a public institution with a research lab embedded within a corporate environment. The agreement aims to lay the foundation for a new research entity that will combine science, the arts, public engagement, and business in a single conceptual space. The goal is to drive innovation and achieve the Sustainable Development Goals (SDGs).

Specifically, CREF and Sony CSL have agreed to work on the following three joint projects:

Sustainable Cities

The inevitable changes our cities are undergoing require smart and strategic thinking to meet the SDGs. To address these challenges, this project aims to develop and implement a series of “what-if” technologies with several specific purposes:

• (i) To evaluate the state of urban environments in terms of public transport quality, inclusivity, and economic activity.

• (ii) To design and test scenarios for urban environments and their socioeconomic challenges.

• (iii) To create metrics, visualizations, modeling tools, and AI assistants that could be adopted by institutional agencies and policymakers worldwide to plan local interventions and reshape the urban fabric.

Disinformation and Social Dialogue (Infosphere)

In this area, JICS aims to improve the health of information dynamics in democratic societies along the following axes:

• (i) To provide a deeper understanding of the supply and demand dynamics of information in the news ecosystem to minimize disinformation and reduce polarization.

• (ii) To develop new recommendation systems and reputation mechanisms to rebuild trust in the information system.

• (iii) To create innovative, scientifically-based policies for governments, authorities, and stakeholders in the news ecosystem.

AI and Machine Learning for Innovation (Augmented Creativity)

JICS intends to study the interactions between Human Intelligence (HI) and Artificial Intelligence (AI) to foster the emergence of what we can call “Augmented Creativity“—human creativity supported by new AI tools. The synthesis of HI and AI, achieved through Augmented Creativity, opens up new scenarios for developing novel ideas and technologies to improve humanity and find sustainable solutions to current social challenges. A side effect of this nascent process will also be a global rethinking of what we currently call Artificial Intelligence, with the ultimate goal of fostering the birth of a new generation of artificially intelligent agents capable of supporting human cognitive endeavors.

Sustainable Cities

The 15-Minute City

The concept of the 15-minute city is a compelling solution for the ecological transition. By reducing daily commutes, we can cut emissions and improve the quality of life. In this context, the JICS initiative provides tools to analyze the current state of cities and develop new sustainable growth scenarios. JICS has enhanced the platform dedicated to 15-minute cities (https://whatif.sonycsl.it/15mincity/), which was initially developed by CSL in Paris, making it more advanced and accessible. The research was published in Nature Cities and was featured on the cover of the dedicated issue. The JICS group is continuing its research on this urban model to understand its implications for sustainability, mobility, and the socioeconomic fabric.

Urban Fitness and Economics

JICS applies the Fitness and Complexity framework to evaluate accessibility and mobility in different urban districts. This approach, initially developed to measure the economic competitiveness of countries through the complexity of their exported products, is being adapted for cities, large metropolitan areas, or regions and their patent production, which serves as a proxy for innovation. A further approach is used to analyze the “fitness landscape” of access to services in cities. This method involves dividing a city into sectors and connecting each sector with the various points of interest (POIs) within it. The goal is to compare the results of this methodology with other urban indicators, such as mobility, accessibility, and real estate value, in order to create a more effective tool for quantifying inequalities and urban segregation phenomena.

Smart-City Project

CREF and Sony CSL, in collaboration with Sony Semiconductor Solutions (SSS) and Sapienza University of Rome, are advancing an innovative project based on the most advanced computer vision technologies applied to an urban context. The pilot project leverages the new Sony IMX500 smart sensor to collect and analyze real-time data. The main activities, carried out in synergy with public bodies and private partners, include:

• Analyzing and forecasting traffic and pedestrian flows in the city while ensuring citizen privacy by design.

• Analyzing road safety at intersections, considering parameters such as the number and speed of vehicles, as well as the interactions between vulnerable users (pedestrians, cyclists) and heavy vehicles, with natural implications for accident prevention and the improvement of urban safety.

The Information Ecosystem (Infosphere)

Understanding and Countering Distortions in the Information Ecosystem

The JICS project aims to study the dynamics of the information ecosystem to understand and counteract the spread of disinformation, the reinforcement of polarization, and the deterioration of public debate. These distortions are the result of biases that influence both the production and consumption of information. On one hand, agenda-setting leads the media to select certain topics and highlight specific perspectives, thus shaping the public’s perception of social priorities. However, when the information supply becomes rigid in predetermined positions without responding to the real demand for information, it creates a space that disinformation can exploit to spread more easily.

On the other hand, confirmation bias drives users to consume and share information that confirms their pre-existing beliefs, reducing their exposure to alternative perspectives and contributing to the fragmentation of public debate. These mechanisms are further amplified by recommendation systems which, by optimizing content visibility based on users’ past interactions, tend to favor the repetition of pre-existing information patterns, limiting the diversity of sources and narrowing the plurality of available information.

Our goal is to develop models capable of anticipating the risks associated with these distortions and evaluating the benefits of strategies that favor pluralistic information and dialogue between different positions, mitigating the toxicity of the debate without compromising freedom of expression.

Solutions for More Equitable, Transparent, and Inclusive Information

Even minimal changes to recommendation system algorithms can alter the flow of information and significantly influence the debate, not only within social platforms but also in traditional media. We are studying strategies to improve the transparency and fairness of these systems, developing metrics capable of measuring and increasing information diversity, going beyond the limitations of current metrics based solely on popularity and engagement. We are also analyzing the costs of polarization, which are not limited to the fragmentation of the debate but also include the degradation of communication quality itself: the reduction of linguistic complexity and the rigidity of digital communities limit the ability to build articulated narratives and foster dialogue between different perspectives. For this reason, we are exploring solutions that can incentivize diversity in content exposure and stimulate more constructive interactions between users with different opinions, helping to make public debate more inclusive and open.

The ultimate goal of our research is to translate these analyses into concrete actions, proposing tools and strategies for digital platforms, policymakers, and journalism, with the aim of improving access to quality information, reducing the negative effects of echo chambers, and promoting an information environment that strengthens democracy rather than undermining it.

AI and Machine Learning for Innovation (Augmented Creativity)

Non-Stationary AI

Modern machine learning is based on the assumption of stationarity, meaning that the parameters of the distributions associated with the training data remain unchanged over time. While it is clear that stationarity rarely occurs in real life, this is especially true for creative processes, for which novelty is a crucial ingredient that can lead to radical paradigm shifts in the reference system. Even the most recent and successful AI systems, such as ChatGPT, do not account for the management of innovation dynamics. JICS will further develop a new method called Dreaming Learning, originally developed by Sony-CSL, to effectively incorporate spatiotemporal novelty into AI systems. JICS intends to deepen and expand this technique through a more profound theoretical understanding of non-equilibrium AI systems and to apply this methodology to real-world problems.

Movement and Innovation with AI

Movement is a means of expression, but also a complex manifestation that combines physical, emotional, and cultural dynamics. What does it mean to interpret or create through movement? Is it possible for a machine to grasp this complexity and contribute to expanding its creative boundaries? These questions are not purely theoretical; they are intertwined with the practical challenges of designing AI systems capable of interacting with the expressive and functional dimensions of human movement. In this context, the JICS initiative is dedicated to exploring movement as a universal language and a means of connection between the natural and the artificial. JICS aims to develop a Large Movement Model, a neural network-based system designed to observe, understand, and predict human movements. The initial focus is on dance, a performing art that represents a particularly rich and complex form of human movement, intertwining rhythm, expression, and creativity. Starting with dance, the model uses the innovative “Dreaming Learning” technique, developed by Sony CSL – Rome, which allows the system to generate original movements, fostering a creative collaboration between dancers, choreographers, and technology. Dance, in this sense, becomes not only an ideal testing ground but also a source of inspiration for exploring the expressive and transformative potential of movement in a broader sense. The applications of this model extend far beyond dance: from physical rehabilitation, where AI can help design personalized exercises, to sports coaching, to the enhancement of performing arts and the preservation of cultural heritage related to human movement. JICS aims to develop a general framework for improving movement quality, making it accessible to professionals in various fields and promoting an innovative dialogue between technology and human capabilities.