From chatbots answering tax questions to algorithms driving autonomous vehicles and handing out medical diagnoses, artificial intelligence supports many aspects of everyday life. According to researchers at the University of California, Irvine, developing smarter, more accurate systems requires a hybrid human-machine approach. In a study recently published in the Proceedings of the National Academy of Sciences, they present a new mathematical model that can improve performance by combining human and algorithmic predictions and confidence scores.

“Human and machine algorithms have complementary strengths and weaknesses. Everyone uses different sources of information and strategies to make predictions and decisions,” said co-author Mark Steyvers, UCI Professor of Cognitive Science. “We show through empirical demonstrations as well as theoretical analysis that humans can improve AI’s predictions even when human accuracy is slightly below [that of] AI – and vice versa. And that accuracy is higher than combining predictions from two people or two AI algorithms.”

To test the framework, researchers ran an image classification experiment in which human participants and computer algorithms worked separately to correctly identify distorted images of animals and everyday objects — chairs, bottles, bikes, trucks. Human participants rated their confidence in the accuracy of each image identification as low, medium, or high, while the machine classifier generated a continuous score. The results showed massive differences in trust between humans and AI algorithms.

“In some cases, human participants were quite confident that a particular image contained a chair, for example, while the AI ​​algorithm was confused about the image,” said co-author Padhraic Smyth, UCI Chancellor Professor of Computer Science. “Similarly, with other images, the AI ​​algorithm was able to reliably provide a label for the object shown, while human participants were unsure whether the distorted image contained a recognizable object.”

When human and AI predictions and confidence scores were combined, the hybrid model resulted in better performance than human or machine predictions obtained alone.

This work breaks new ground by demonstrating the potential of combining human and machine predictions and pointing to new and improved approaches for human-AI collaboration,” said Smyth.

This study will provide further insights into how humans and machines can work together to build more accurate artificially intelligent systems.

What is meant by artificial intelligence?

Artificial intelligence (AI) or machine intelligence refers to computers and robots capable of analyzing data in context to provide requested information and analysis, or to trigger events based on results. Through techniques like machine learning and neural networks, companies around the world are investing in teaching machines to “think” more like humans.

Dlab Innovations tracks the latest research and technology developments in the field of artificial intelligence, including the companies, applications, innovations and platforms that will shape the future of AI.