The project conducts research on new conversational AI technologies that understand a user’s mental health conditions and enable a principled strategy to counsel the user. The research will focus on incorporating user personalisation and counselling strategies into the AI models. At the end of project, we hope to create a conversational AI framework that can automate mental health counselling and evaluate its performance.
In the dynamic field of wearable health technology, our proposed research aims to revolutionise how we monitor our health using devices such as smartwatches and earbuds. These devices frequently employ photoplethysmography (PPG), a noninvasive technique that monitors changes in blood volume under the skin, providing valuable insights into cardiovascular health. However, real-world challenges, such as inaccuracies during physical activities and the impact of diverse body postures, impede the realisation of the full potential of PPG technology with respect to these wearable devices.
Our research focuses on a breakthrough concept: incorporating contact pressure (CP) into PPG measurements to address the aforementioned challenges. By analyzing the tightness of wearable devices against the skin, we aim to obtain valuable insights that can help reconstruct high-quality PPG data from the noisy PPG data. Our first contribution is the development of a wearable prototype capable of concurrently measuring CP and PPG. Using this prototype, we will develop intelligent algorithms to mitigate the effects of physical activities and body postures. Finally, we will optimise our energy efficiency and real-time processing methodologies, ensuring prolonged battery life for wearable devices.
We believe that these innovations can substantially improve the accuracy and reliability of health data obtained from wearables, thereby unlocking new capabilities of PPG in health monitoring. The success of this research has the potential to stimulate market growth by establishing a new standard for accuracy and capabilities in wearable devices. Considering the global aging population, our research will considerably impact elderly care, particularly cardiovascular diseases. By improving the reliability of wearable devices, our research can promote an active lifestyle and contribute to overall well-being among the general population.
Artificial intelligence (AI)-driven personal assistants like Apple’s Siri and Amazon’s Alexa have gained significant popularity due to their user-friendly natural-language interfaces that facilitate interaction with the Internet. These agents, embedded in various devices, assist users with tasks such as retrieving information, controlling home automation, and managing calendars. However, current assistants lack awareness of the physical environment; they cannot connect directly to sensors to interpret a user's physical environment, limiting their ability to respond to contextual queries, such as identifying the manufacturer of a toaster pointed out by a user.
Recent advances in AI, particularly with models like GPT-4, have enhanced capabilities for machine-based reasoning that combine visual, verbal, and gestural inputs. Projects like Google’s Project Astra are exploring these multi-modal assistants, which could revolutionize human-computer interactions. Additionally, the integration of sophisticated sensors in mobile and IoT devices, such as LIDAR technology in smartphones, opens the door for new situated agents that can leverage spatial awareness and multimodal reasoning to deliver contextually relevant information.
Developing these situated agents presents significant challenges, notably in energy consumption and computational complexity. High-power sensors like LIDAR are unsuitable for continuous use, and current AI models are too large to run efficiently on mobile devices. To address these issues, the project proposes innovations in energy-efficient sensing techniques and optimization of deep neural networks (DNNs) to minimize latency and energy use. Through innovations in triggering novel spatial sensors on-demand, processing relevant multi-modal sensor data selectively and by exploiting the cached results of recent reasoning actions, these advancements aim to enable responsive and pervasive situated agents that can interact meaningfully with users. The project also seeks to embed these innovations in a prototype conversational agent that can serve as a real-time tutor, responding intelligently to a learner’s queries and enhancing educational experiences.
In recent years, the rapid advancement of Large Language Models (LLMs) has driven significant progress in Generative Artificial Intelligence (GAI). By emulating human language capabilities, GAI has unlocked a multitude of applications, ranging from chatbots and virtual assistants to translation services and content-generation tools. However, GAI has evolved into a double-edged sword, giving rise to ethical concerns surrounding transparency, privacy protection, misinformation, bias and fairness, job displacement, and environmental impact.
In this project, the team aims to quantify and validate firms' ethical awareness of GAI and employ econometrics models and quasi-experiments to comprehend its determinants and document consequences. Through these efforts, the project makes a distinctive contribution to the promising trajectory of developing responsible and ethical GAI systems in business, ultimately fostering a sustainable society.
The research aims to leverage modern Large Language Models (such as ChatGPT) to augment scientific practices for causal inference. Consumer behavior plays a critical role in shaping the economy, yet it is dynamically evolving and difficult to predict. The project seeks to equip social scientists with cutting-edge AI-based toolkits to enhance the study of consumer behavior in today's rapidly changing marketplaces, thereby improving understanding for researchers, businesses, and policymakers.
This research investigates the role of AI in enhancing human creativity across three stages: idea generation, idea selection, and idea evaluation. It aims to better understand how collaboration between humans and AI can improve creativity by increasing the quantity and diversity of ideas and reducing the time required for creative tasks. The study will also examine if AI offers greater creative advantages to women and individuals in rule-abiding societies who face constraints due to social norms.
Despite the increased importance of environmental issues for firms, environmental compliance costs are high because of an inefficient and ineffective regulatory system, compounded by the reluctance of businesses to act proactively. Although many firms often recognize the costs of environmental issues for external stakeholders, they may overlook the impact of these issues on internal operations. This project investigates how a firm’s environmental performance impacts its approach toward human capital, including information management, employee retention, recruitment, and talent quality. Additionally, it aims to examine the potential negative externalities of corporate environmental incidents on the information environment in labor markets.
High frequency financial data provides useful real time information about the global economy, represented by stocks, bonds, futures, derivatives, exchange rates, and crypto currencies. Such information is important not only for institutional investors and regulators, but also for ordinary citizens who participates in financial markets. This project aims to develop new econometric methods to optimally extract information from high frequency financial data and investigate how to conduct robust statistical inference concerning economic hypotheses.
Climate change is an increasingly urgent and existential crisis that threatens all sectors of society in all countries of Southeast Asia. The Climate Transformation Programme aims to develop, inspire and accelerate knowledge-based solutions and educate future leaders to establish the stable climate and environment necessary for resilient, just, and sustainable Southeast Asian societies. CTP will generate knowledge and innovation across disciplines including climate and Earth science, ecology, materials science, artificial intelligence, humanities, social sciences and the arts, finance, health, and engineering. It will also translate state-of-the-art scientific results into real world solutions for Singapore, and it will transfer these solutions to Southeast Asia and beyond.
CTP will help to better understand exactly how the climate is changing; to identify and develop ways to contribute to global climate change mitigation efforts; to better assess the environmental, social, and economic risks from climate change; and to design effective solutions to control, reduce and adapt to these risks.
To achieve these multiple objectives, CTP will integrate a matrix of six strategic clusters of research and three cross-cutting themes, where Sonny Rosenthal will participate in and contribute to Cross Cutting Theme 1: Sustainable societies.
(Cross Cutting Theme 1: Sustainable societies will leverage climate research in communication, public opinion and psychology, and artistic impressions to shape public attitudes and behaviours.)
The project aims to enhance the understanding of the legal and economic aspects of sustainability-related trade, investment, and financial relations between the Association of Southeast Asian Nations (ASEAN) countries and Taiwan.