Artificial intelligence (AI) and machine learning (ML) are rapidly transforming various sectors, and their potential to address complex social problems is becoming increasingly apparent. This guide explores how machine learning can be leveraged to tackle issues ranging from mental health and accessibility to environmental concerns and public health crises.
Introduction
The rise of AI has been marked by impressive milestones, such as IBM’s Watson winning Jeopardy! and Google’s AlphaGo defeating the world Go champion. These achievements, along with the development of self-driving cars, highlight the transformative power of AI. As countries and corporations invest heavily in AI research and development, the application of these technologies to social problem-solving is gaining momentum.
IBM’s Science for Social Good initiative exemplifies this trend, using AI to address challenges in healthcare, education, and environmental protection. One notable project involved using machine learning to understand and predict the spread of the Zika virus, enabling targeted surveillance and management efforts [1].
While investments in AI often focus on industrial and service growth, there’s a growing recognition of the need to allocate more resources to address social issues. In South Korea, for instance, the government is increasing its investment in ICT (Information and Communication Technology) for social problem-solving, with AI playing a key role in improving quality of life and addressing population growth [2].
This guide introduces research on an informatics platform for social problem-solving, based on spatio-temporal data. This platform aims to develop methodologies that can explain, predict, and address diverse social problems through a convergence of social sciences, data science, and AI.
Pilot Research and Studies on Social Problem Solving
To examine the applicability of AI and machine learning in addressing various social problems, several pilot studies were conducted, focusing on:
- Analysis of individual characteristics related to suicidal impulses
- Study on the mobility of disabled individuals using GPS data
- Visualization of anxiety distribution using Social Network Services (SNS)
- Big data-based analysis of noise environments and exploration of solutions
- Analysis of the response governance regarding the Middle Eastern Respiratory Syndrome (MERS)
These studies explore research issues, test the validity of convergent-scientific methodologies, and examine the potential for resolving these problems. The collected data and information are stored in a knowledge base (KB), along with research methods used in data extraction, collection, analysis, and visualization. This KB and method database are then merged into an open informatics platform.
Analysis of Individual Characteristics with Suicidal Impulse
South Korea faces a significant challenge with a high suicide rate compared to other OECD countries (Fig. 1). Understanding the causes of suicide is crucial for establishing effective prevention measures. This research aims to understand suicidal impulses by analyzing the characteristics of individuals who have experienced them, predict the likelihood of suicide attempts, and establish policies to help prevent suicide.
Fig. 1.
Suicide rates in OECD countries for 2013, highlighting the high rate in South Korea.
Using data mining techniques on the Korean Social Survey and Survey of Youth Health Status Data, researchers analyze suicide risk groups. They employ a predictive model based on cell propagation to overcome the limitations of traditional statistical methods. This approach identifies correlations between suicide impulses and individual attributes, such as gender, age, education, marital status, level of satisfaction, disability status, occupation status, housing, and household income.
The analysis identifies high-risk suicide clusters, with one group (C1) characterized by low income and education levels (Fig. 2). The level of satisfaction with life, disability, marital status, and housing have the highest impact on suicidal impulse. Women and young people also tend to have a higher risk.
Fig. 2.
Visualization of suicide risk groups represented by household income and level of education.
Future research involves developing new prediction models with other machine learning methods and establishing mitigation policies. Subjective analyses of well-being, social exclusion, and spatio-temporal characteristics will also be explored.
Study on the Mobility of the Disabled Using GPS Data
Mobility rights are fundamental to quality of life for all individuals, including those with disabilities. This study aims to suggest policies that extend mobility rights for the disabled by analyzing travel patterns and socio-demographic characteristics of individuals with physical impairments.
Traditional studies often rely on travel diaries, interviews, and questionnaire surveys. This research uses geo-location tracking GPS data collected via mobile devices to analyze mobility patterns (distance, speed, frequency of outings) through regression analysis.
Participants with physical disabilities used an open mobile application called traccar to collect GPS data over a month. The resulting trajectories are visualized, providing insights into their mobility patterns (Fig. 3).
Fig. 3.
Visualization of the trajectory of disabled individuals using geo-location data.
GPS data offers more detailed insights into mobility status compared to conventional questionnaire surveys. The analysis revealed that the disabled often prefer bus routes that visit diverse locations over the shortest route. Age and monthly income are negatively associated with mobility.
Based on these findings, the study suggests developing new bus routes for the disabled and recommending a new location for the current welfare center to enable a greater range of travel. Further research will explore travel patterns using indoor positioning technology and CCTV image data.
Visualization of the Distribution of Anxiety Using Social Network Services
Social anxiety underlies many pressing social issues, including political polarization, competition in education, and rising suicide rates. Increased social anxiety can intensify competition and conflict, hindering social solidarity and decreasing trust.
This research uses the Internet and social media to access emotional traits, leveraging platforms for information exchange and emotional response sharing. By capturing emotional responses and geo-locations in real-time through machine learning, the study aims to visualize the spatio-temporal distribution of anxiety.
A visualization system was built to map the regional and temporal distribution of anxiety by combining spatio-temporal information from Twitter with sentiment analysis. A Twitter message collecting crawler was developed to build a dictionary and tweet corpus. This enabled the development of an automatic classification system for anxiety-related messages, visualizing the nationwide distribution of anxiety (Fig. 4).
Fig. 4.
Illustration of the Twitter message classification process.
A Naïve Bayes Classifier was used for anxiety identification, achieving an accuracy of 84.8%. The system revealed regional disparities in anxiety emotions. It also found that Twitter users in politicized regions tend to disclose less about their residing areas.
The Twitter-based system can continuously classify accumulated tweet text data and provide a temporal visualization of anxiety distribution at various scales (by ward, city, province, and nationwide) (Fig. 5), compensating for the limitations of traditional survey methods.
Fig. 5.
Visualization of the regional distribution of anxiety in Korean society by geo-scale.
Big Data-Based Analysis of Noise Environment and Exploration of Technical and Institutional Solutions for Its Improvement
Environmental issues are a major social concern, with growing interest in the effects of environmental aesthetics on quality of life. While there’s significant effort to improve the visual environment, less attention has been given to the auditory environment.
This study aims to provide a cognitive-based, human-friendly solution to improve noise problems by developing a tool for collecting sound data and converting it into a sound database. It also seeks to build spatio-temporal features and a management platform for indoor and outdoor noise sources.
Pilot experiments were conducted to predict indicators that measure emotional reactions by developing a handheld device application for data collection. Participants were asked to indicate whether sounds they heard were “Good” or “Bad”, triggering a recording of the sound and a series of questions about the location and auditory environment. The subject’s paths can be displayed to analyze the relations of the soundscapes to the paths (Fig. 6).
Fig. 6.
Subject’s paths and marks for sound types.
The study helped to build a positive auditory environment for specific places, provide policy data for noise regulation, and identify areas that are alienated from the auditory environment.
Analysis of the Response Governance Regarding the Middle Eastern Respiratory Syndrome (MERS)
The spread of new infectious diseases is a growing concern, with epidemics having profound social and economic impacts. Establishing an effective rapid response system (RRS) for infectious disease control is essential.
This study compares the official response system with the actual response system during the MERS outbreak in Korea in 2015. It aims to understand the institutional mechanism of the epidemic response system and identify effective policy alternatives.
An automatic news article crawling tool was developed to collect and analyze web-based newspaper articles related to MERS. The tool extracted information into triplet graphs (subjects/verbs/objects) from the articles using natural language processing techniques.
A tool for visualizing information at a specific time with a network graph was developed and utilized to facilitate analysis and visualization of the networks (Fig. 8).
Fig. 8.
Visualization of a graph network depicting the MERS response at a specific time.
The analysis revealed discrepancies between the official crisis response manual and the actual response during the MERS outbreak. The study suggests new policies for infectious disease control, including strengthening cooperation networks in early response systems and developing more effective management plans.
Convergent Approaches and Open Informatics Platform
Addressing social issues often requires obtaining evidence from massive datasets, which can be challenging for traditional social sciences. Data science and AI can support social science by discovering hidden contexts and patterns through low-cost analyses of large data. However, knowledge derived solely from machine learning can lack validity.
Social science can help data science and AI by interpreting social phenomena, verifying theoretical validity, and identifying causal relationships. This necessitates convergent-scientific approaches for social problem-solving.
These convergent approaches offer the possibility of building an informatics platform that can interpret, predict, and solve various social problems through the combination of social science and data science (Fig. 9). The platform integrates data sets and tools for data collection, analysis, and visualization.
Fig. 9.
Diagram illustrating the structure of the informatics platform.
The Open Informatics Platform is planned to be expanded to incorporate citizen sensing, in which people’s observations and views are shared via mobile devices and Internet services in the future [7].
Conclusions
Solving social problems requires both technical and social approaches. An integrated marriage between the two is essential for effective social problem management. It’s crucial to carefully define indicators specific to a given social problem, recognizing that many qualitative indicators cannot be directly measured [8].
While machine learning technologies offer cost-effective solutions, they can introduce bias and errors. Human experts are needed to recognize and correct erroneous outputs and interpretations [10]. New AI frameworks that can infer and recognize objects based on small amounts of data, such as Transfer Learning [11], generate lacking data (GAN), or integrate traditional AI technologies are worth exploring.
Although AI is progressing rapidly, challenges remain in solving social problems. Further research on the integration of social science and AI is required. AI should primarily be used as a decision aid, considering its current limitations and capabilities.
Acknowledgements
This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korean government (MSIT1) (No. 2018R1A5A7059549).
Footnotes
1Ministry of Science and ICT.
Contributor Information
Fernando Koch, Email: [email protected].
Atsushi Yoshikawa, Email: [email protected].
Shihan Wang, Email: [email protected].
Takao Terano, Email: [email protected].
Yeunbae Kim, Email: [email protected].
Jaehyuk Cha, Email: [email protected].
References
- 1.IBM: Science for Social Good – Applying AI, cloud and deep science toward new societal challenges. http://www.research.ibm.com/science-for-social-good/
- 2.MSIT of Korea: I-KOREA 4.0 ICT R&D Innovation Strategy (2018)
- 3.Kim, N., Hong, J., Kim, H., Kim, S.: Analyzing suicide-ideation survey to identify high-risk groups: a data mining approach. In: International Conference on Green and Human Information, China (2017)
- 4.Kim, H., Lee, Y., Cha, J.: Mobility among people with physical impairment: a study using geo-location tracking data. In: International Conference on Green and Human Information, China (2017)
- 5.Lee, J., Kim, J., Choi, Y.: SNS data visualization for analyzing spatial-temporal distribution of social anxiety. In: EDB 2016 Proceedings of the Sixth International Conference on Emerging Databases: Technologies, Applications, and Theory, Jeju, Korea, pp. 106–109 (2016)
- 6.Kim, J., Jung, J., Cha, J., Choi, J., Choi, C., Oh, S.: Application of network analysis into emergency response: focusing on the 2015 outbreak of the middle-eastern respiratory syndrome in Korea. In: Information, vol. 21, No. 2, pp. 441–446. International Information Institute, Tokyo (2018)
- 7.Koch F, Cardonha C, Gentil JM, Borger S. A platform for citizen sensing in sentient cities. In: Nin J, Villatoro D, editors. Citizen in Sensor Networks; Heidelberg: Springer; 2013. pp. 57–66. [Google Scholar]
- 8.Land, K.C., Ferriss, A.L.: Chapter 52: The sociology of social indicators. In: 21st Century Sociology (2006). http://www.soc.duke.edu/~cwi/Section_I/I-19TheSociologyofSocialIndicators.pdf
- 9.Pentland A. To signal is human. Am. Sci. 2010;98(3):204–211. doi: 10.1511/2010.84.203. [DOI] [Google Scholar]
- 10.Kleinberg, J., Ludwig, J., Mullainathan, S.: A Guide To Solving Social Problems With Machine Learning. In: Harvard Business Review (2016). https://hbr.org/2016/12/a-guide-to-solving-social-problems-with-machine-learning
- 11.Yosinski, J., Clune, J., Bengio, Y., Lipson, H.: How transferable are features in deep neural networks? In: Advances in Neural Information Processing Systems 27 (NIPS 2014). NIPS Foundation (2014)
