Human-AI collaboration improves source search outcomes

When artificial intelligence robots that have been designed to use algorithms to complete source search tasks, such as search and rescue operations during a fire, encounter a disturbance, they are often unable to complete their task. Proposed solutions have ranged from trying to improve algorithms to introducing additional robots, but these AI-driven robots still encounter fatal problems.

Researchers have proposed a solution: a human-AI collaboration that takes advantages of the unique skills of the human brain to overcome challenges.

The paper was published in the Journal of Social Computing in June, 2023.

“It is time to bring humans back,” said Yong Zhao, a researcher from Changsha, China. “AI-driven robots are often used in situations when physical search would be too dangerous or physically impossible for people, such as locating the origin of a fire or identifying the source of toxic gas. However, AI robots can encounter critical problems that cannot be resolved autonomously, such as getting stuck or misidentifying the source. These are problems that are easily tackled by humans using their expertise, experience, and even instincts. A crowd-powered system offers a novel solution.”

To prove the feasibility of their human-AI collaboration strategy, researchers first identified the different types of hazards the robots could encounter. These hazards were then sorted into whether or not a human observer could help AI solve the problem. If the problem cannot be solved with human assistance, for example, if the search area is too large, then the search is stopped. However, if the problem could be solved with human assistance, the AI develops an explanation of the problem, and it is sent for crowdsourcing. 

“Involving humans in the automated problem-solving process enhances the efficacy and efficiency of the algorithm. In scenarios where the robot faces challenges due to dynamic, deteriorated, or unfamiliar environments, temporary human intervention can be employed without prior knowledge of the surroundings to address these issues. Once resolved, AI seamlessly resumes control over the robot to continue its search,” said Sihang Qiu.

After identifying the different types of hazards and whether or not humans could assist in source search scenarios, researchers developed a user study. The user study tested two different control modes of the AI robot—Full Control and Aided Control. In Full Control, the human collaborator takes over the search process. In Aided Control, the problem-solving decision tree decides if the human-AI collaboration would be beneficial.

During Aided Control, when they received information from the algorithm about the problem and did not give over complete control, participants felt like they had less cognitive workload and could address the problem. However, non-experts had a harder time understanding the AI-driven robot’s explanations of the problem, leading the researchers to propose personalized interactions based on the experience of the human in the collaboration, including plain language explanations.

Looking ahead, researchers will try to find ways to include additional personalization, based on the human participants’ background, education level, and personality. “This study paves the way for our future exploration into harnessing crowd-powered systems to facilitate effective collaboration between humans and AI. Our aim is to substantiate the manifold advantages of such collaboration in diverse application scenarios, including but not limited to natural language processing and image analysis,” said Qiu.

Other contributors include Zhengqiu Zhu and Bin Chen from Changsha, China.

The National Natural Science Foundation of China and the Postgraduate Scientific Research Innovation Project of Hunan Province funded this research.

##

About Journal of Social Computing

Journal of Social Computing (JSC) is an open access, peer-reviewed scholarly journal which aims to publish high-quality, original research that pushes the boundaries of thinking, findings, and designs at the dynamic interface of social interaction and computation. This will include research in （1）computational social science—the use of computation to learn from the explosion of social data becoming available today; (2) complex social systems or the analysis of how dynamic, evolving social collectives constitute emergent computers to solve their own problems; and (3) human computer interaction whereby machines and persons recursively combine to generate unique knowledge and collective intelligence, or the intersection of these areas. The editorial board welcomes research from fields ranging across the social sciences, computer and information sciences, physics and ecology, communications and linguistics, and, indeed, any field or approach that can challenge and advance our understanding of the interface and integration of computation and social life. We seek to take risks, avoid boredom and court failure on the path to transformative new paradigms, insights, and possibilities.  The journal is open to a diversity of theoretic paradigms, methodologies and applications.

About SciOpen

SciOpen is a professional open access resource for discovery of scientific and technical content published by the Tsinghua University Press and its publishing partners, providing the scholarly publishing community with innovative technology and market-leading capabilities. SciOpen provides end-to-end services across manuscript submission, peer review, content hosting, analytics, and identity management and expert advice to ensure each journal’s development by offering a range of options across all functions as Journal Layout, Production Services, Editorial Services, Marketing and Promotions, Online Functionality, etc. By digitalizing the publishing process, SciOpen widens the reach, deepens the impact, and accelerates the exchange of ideas.