# Human-AI Symbiosis: A Survey of Current Approaches

**Authors:** Zahra Zahedi, Subbarao Kambhampati  
**arXiv:** [2103.09990](https://arxiv.org/abs/2103.09990)  
**Date:** 2021-03-18  
**Summary page:** https://www.summarizepaper.com/en/paper/2103.09990v1/

## Key points

- The paper provides a comprehensive outline of different threads of work in human-AI collaboration
- Various aspects of works on the human-AI team are highlighted, including complementing flow, task horizon, model representation, knowledge level, and teaming goal
- The authors create a taxonomy of recent works in the field based on these dimensions
- The survey serves as a valuable resource for researchers interested in human-AI collaboration
- Understanding how humans and AI systems can effectively collaborate to achieve common goals is emphasized
- Different approaches to the flow of complementing are explored
- Short-term and long-term tasks are considered within the task horizon dimension
- Model representation techniques such as rule-based systems or machine learning algorithms are discussed
- The knowledge level dimension examines how humans and AI systems share information and expertise
- Teaming goals include improving efficiency, enhancing decision-making processes, or achieving better outcomes through collective intelligence.

## Summary

In their paper titled "Human-AI Symbiosis: A Survey of Current Approaches," Zahra Zahedi and Subbarao Kambhampati aim to provide a comprehensive outline of the different threads of work in human-AI collaboration. They highlight various aspects of works on the human-AI team, including the flow of complementing, task horizon, model representation, knowledge level, and teaming goal. By analyzing these dimensions, the authors create a taxonomy of recent works in the field. The survey serves as a valuable resource for researchers interested in human-AI collaboration by offering a clear connection between existing works and providing guidance for future research endeavors. The authors emphasize the importance of understanding how humans and AI systems can effectively collaborate to achieve common goals. By examining the flow of complementing, which refers to how humans and AI systems exchange information and support each other's tasks, the authors shed light on different approaches that have been explored. They also consider the task horizon, which encompasses short-term tasks that require immediate attention as well as long-term tasks that involve planning and decision-making over extended periods. Another dimension explored is model representation, which focuses on how AI systems represent knowledge and make it accessible to humans. The authors discuss various techniques employed in representing models such as rule-based systems or machine learning algorithms. Furthermore, the knowledge level dimension examines how humans and AI systems share information and expertise. This includes considering whether AI systems possess domain-specific knowledge or if they can acquire knowledge through interaction with humans. Lastly, the teaming goal dimension addresses the overall objective of human-AI collaboration. The authors categorize goals into areas such as improving efficiency, enhancing decision-making processes, or achieving better outcomes through collective intelligence. Overall, this survey provides an extensive overview of recent advancements in human-AI collaboration. It not only categorizes existing works based on key dimensions but also offers insights into potential future directions for research in this rapidly evolving field by highlighting important considerations for successful collaborations between humans and AI systems.