[Paper Review] Artificial Intelligence for Social Good: A Survey

Zheyuan Ryan Shi, Claire Wang+1 more|arXiv (Cornell University)|Jan 7, 2020

COVID-19 and Mental HealthPsychology343 references52 citations

TL;DR

This survey analyzes the AI for Social Good (AI4SG) literature, quantifies trends across eight domains and AI techniques, and proposes a unified framework (AEC and DPP) to categorize AI4SG work while highlighting future research directions.

ABSTRACT

Artificial intelligence for social good (AI4SG) is a research theme that aims to use and advance artificial intelligence to address societal issues and improve the well-being of the world. AI4SG has received lots of attention from the research community in the past decade with several successful applications. Building on the most comprehensive collection of the AI4SG literature to date with over 1000 contributed papers, we provide a detailed account and analysis of the work under the theme in the following ways. (1) We quantitatively analyze the distribution and trend of the AI4SG literature in terms of application domains and AI techniques used. (2) We propose three conceptual methods to systematically group the existing literature and analyze the eight AI4SG application domains in a unified framework. (3) We distill five research topics that represent the common challenges in AI4SG across various application domains. (4) We discuss five issues that, we hope, can shed light on the future development of the AI4SG research.

Motivation & Objective

• Summarize the AI4SG research landscape and its growth over 2008–2019.
• Quantitatively analyze the distribution of AI4SG literature across eight application domains and AI techniques.
• Propose three conceptual methods to systematically group AI4SG literature within a unified framework.
• Distill five cross-domain AI research topics and discuss deployment/evaluation challenges.
• Provide a case study and discuss future directions for AI4SG research.

Proposed method

• Selected 1,176 papers from AAAI, IJCAI, IAAI, AAMAS, KDD, ACM SIGKDD/COMPASS conferences (2008–2019).
• Tagged papers with eight domains and sixteen AI sub-topics using keyword-based and manual curation; refined with iterative keyword matching.
• Developed three grouping approaches: (i) domain/topic structure, (ii) agent–environment–community (AEC) scope, (iii) Descriptive–Predictive–Prescriptive (DPP) AI intervention functionality.
• Analyzed temporal trends and produced a domain–techniques heatmap to identify hot areas and combinations (e.g., ML in healthcare).
• Provided domain-specific descriptions and a case study (Kudu market in Uganda) to illustrate deployment and retrospective insights.

Experimental results

Research questions

• RQ1 What are the temporal trends and domain distribution of AI4SG research from 2008 to 2019?
• RQ2 Which AI techniques are most prevalent within AI4SG, and how do they vary by domain?
• RQ3 How can AI4SG literature be systematically categorized across domains, agent scope, and AI intervention functionality?
• RQ4 What common challenges and opportunities emerge across AI4SG domains, and how do deployment experiences inform future research?
• RQ5 What are representative deployed/piloted AI4SG projects and their lessons learned?

Key findings

• AI4SG publications grew from 18 in 2008 to 246 in 2019 across all domains and techniques.
• Healthcare is the most studied domain, accounting for about 32% of AI4SG literature in 2019, with transportation ranking second.
• Machine learning is the dominant technique across AI4SG, increasingly so since 2013, often underpinning CV and NLP within other domains.
• A heatmap shows ML is central across domains, with transportation, healthcare, public safety, and environmental sustainability being the most studied domains.
• The strongest domain–technique pairing is ML in healthcare (174 papers), illustrating a key cross-domain hotspot; other notable pairings include planning/optimization in transportation and human computation for combating information manipulation.
• The Kudu case study demonstrates on-the-ground deployment challenges, including market design, data needs, and hybrid algorithmic–human clearing.