Human-AI collaboration is the study of how humans and
artificial intelligence (AI) agents work together to accomplish a shared goal.[1] AI systems can aid humans in everything from decision making tasks to art creation.[2] Examples of collaboration include
medical decision making aids.,[3][4] hate speech detection,[5] and
music generation.[6] As AI systems are able to tackle more complex tasks, studies are exploring how different models and explanation techniques can improve human-AI collaboration.
Improving collaboration
Explainable AI
When a human uses an AI's output, they often want to understand why a model gave a certain output.[7] While some models, like decision trees, are inherently explainable, black box models do not have clear explanations. Various
Explainable artificial intelligence methods aim to describe model outputs with post-hoc explanations[8] or visualizations,[9] these methods can often provide misleading and false explanations.[10] Studies have also found that explanations may not improve the performance of a human-AI team, but simply increase a human's reliance on the model's output.[11]
Trust in AI
A human's trust in an AI agent is an important factor in human-AI collaboration, dictating whether the human should follow or override the AI's input.[12] Various factors impact a person's trust in an AI system, including its accuracy[13] and reliability[14]
Adoption of AI
AI adoption by users is crucial for improving Human-AI collaboration since userâs adoption is not just about using the new technology, but also important in transforming how work is done, how decisions are made, and how projects and organizations operate in a more efficient manner. This transformation is essential for realizing the full potential of Human-AI collaboration.
In the evolving digital landscape, there is an increasing pressure to adopt and effectively utilize artificial intelligence (AI), which is steadily entering the management, work, and organizational ecosystems and enabling digital transformations. The successful adoption of AI is a complex and multifaceted process that requires careful consideration of various factors [15]
^Roberts, Adam; Engel, Jesse; Mann, Yotam; Gillick, Jon; Kayacik, Claire; NĂžrly, Signe; Dinculescu, Monica; Radebaugh, Carey; Hawthorne, Curtis; Eck, Douglas (2019).
"Magenta Studio: Augmenting Creativity with Deep Learning in Ableton Live". Proceedings of the International Workshop on Musical Metacreation (MUME).
^Ribeiro, Marco Tulio; Singh, Sameer; Guestrin, Carlos (2016-08-13).
""Why Should I Trust You?"". Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. KDD '16. San Francisco, California, USA: Association for Computing Machinery. pp. 1135â1144.
doi:
10.1145/2939672.2939778.
ISBN978-1-4503-4232-2.
S2CID13029170.
^Bansal, Gagan; Wu, Tongshuang; Zhou, Joyce; Fok, Raymond; Nushi, Besmira; Kamar, Ece; Ribeiro, Marco Tulio; Weld, Daniel S. (2021-01-12). "Does the Whole Exceed its Parts? The Effect of AI Explanations on Complementary Team Performance".
arXiv:2006.14779 [
cs.AI].