‘It’s time to put the humans back’: Human-AI collaboration improves source search results

When artificial intelligence robots designed to use algorithms to complete source-finding tasks, such as search and rescue operations during a fire, encounter disturbances, they often cannot complete them. become your duty. Proposed solutions range from trying to improve algorithms to introducing more robots, but these AI-controlled robots still encounter serious problems.

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

The article was published in the magazine Journal of Social Computing.

“It’s time to bring humans back,” said Yong Zhao, a researcher from Changsha, China.

“AI-controlled robots are often used in situations where a physical search is too dangerous or impossible for humans, such as determining the origin of a fire or identifying the source of toxic gas However, AI robots can encounter serious problems that cannot be solved automatically, such as getting stuck or misidentifying the source. These are problems that humans have can be easily solved with one’s expertise, experience and even instinct. A crowd-powered system offers a new solution.”

To demonstrate the feasibility of a human-AI collaborative strategy, the researchers first identified different types of dangers that robots might encounter. These dangers are then ranked according to whether the observer can help the AI ​​solve the problem. If the problem cannot be solved with human help, for example if the search area is too large, the search will be stopped. However, if the problem can be solved with human help, the AI ​​will provide an explanation for the problem and send it out to provide resources to the community.

“Involving humans in the automated problem-solving process will improve the efficiency and effectiveness of the algorithm. In situations where robots face challenges due to dynamic, degraded, or otherwise unstable environments familiar, temporary human intervention without prior knowledge of the surrounding environment can be used to solve these problems. Once resolved, the AI ​​will continue to control the robot seamless way to continue searching,” Sihang Qiu said.

After determining the different types of hazards and whether humans could assist in source-finding situations, the researchers developed a user study. The user study tested two different control modes of the AI ​​robot—Full Control and Assisted Control. In Full Control, a human collaborator takes over the search process. In Control Support, the problem-solving decision tree determines whether human-AI collaboration is beneficial or not.

During Assisted Control, when they received information from the algorithm about the problem and were not given full control, participants felt like they had less cognitive workload and could solve the problem. solve the problem. However, non-experts had a harder time understanding the AI-driven robot’s explanation of the problem, leading the researchers to recommend personalized interactions based on their experiences. people in the collaborative process, including explanations in plain language.

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