Reciprocal human machine learning

RHML emerged in the context of the rise of big data analytics and artificial intelligence for intelligent tasks like sense-making and decision-making.{{Cite journal |last1=University of Virginia |last2=Abbasi |first2=Ahmed |last3=Zhou |first3=Yilu |last4=Fordham University |last5=Deng |first5=Shasha |last6=Shanghai International Studies University |last7=Zhang |first7=Pengzhu |last8=Shanghai Jiaotong University |date=2018-02-02 |title=Text Analytics to Support Sense-Making in Social Media: A Language-Action Perspective |url=https://misq.org/skin/frontend/default/misq/pdf/appendices/2018/V42I2Appendices/04_13239_RA_AbbasiZhou.pdf |journal=MIS Quarterly |volume=42 |issue=2 |pages=427–464 |doi=10.25300/MISQ/2018/13239}} As machine learning advanced to take on more roles, researchers realized fully autonomous systems had limitations and needed human guidance.{{Cite journal |last1=van den Broek |first1=Elmira |last2=Sergeeva |first2=Anastasia |last3=Huysman |first3=Marleen |date=2021-09-01 |title=When the Machine Meets the Expert: An Ethnography of Developing AI for Hiring |url=https://aisel.aisnet.org/misq/vol45/iss3/21 |journal=Management Information Systems Quarterly |volume=45 |issue=3 |pages=1557–1580 |doi=10.25300/MISQ/2021/16559 |s2cid=238222671 |issn=0276-7783|url-access=subscription }}{{Cite journal |last1=Sturm |first1=Timo |last2=Gerlach |first2=Jin |last3=Pumplun |first3=Luisa |last4=Mesbah |first4=Neda |last5=Peters |first5=Felix |last6=Tauchert |first6=Christoph |last7=Nan |first7=Ning |last8=Buxmann |first8=Peter |date=2021-09-01 |title=Coordinating Human and Machine Learning for Effective Organization Learning |url=https://aisel.aisnet.org/misq/vol45/iss3/22 |journal=Management Information Systems Quarterly |volume=45 |issue=3 |pages=1581–1602 |doi=10.25300/MISQ/2021/16543 |s2cid=238222756 |issn=0276-7783|url-access=subscription }}{{Cite journal |last=Elizabeth |first=Abigail |date=2023-09-08 |title=Reciprocal, human-machine learning could transform how researchers open science |url=https://osf.io/q5my6/ |journal=OSF |language=en}}{{Cite journal |last1=Nixdorf |first1=Steffen |last2=Zhang |first2=Minqi |last3=Ansari |first3=Fazel |last4=Grosse |first4=Eric H. |date=2022-01-01 |title=Reciprocal Learning in Production and Logistics |journal=IFAC-PapersOnLine |series=10th IFAC Conference on Manufacturing Modelling, Management and Control MIM 2022 |volume=55 |issue=10 |pages=854–859 |doi=10.1016/j.ifacol.2022.09.519 |s2cid=253195774 |issn=2405-8963|doi-access=free }}

RHML extends the concept of human-in-the-loop systems by promoting reciprocal learning. Humans learn from their interactions with machine learning models, staying up-to-date on evolving technology.{{Cite book |last=So |first=Chaehan |title=Artificial Intelligence in HCI |chapter=Human-in-the-Loop Design Cycles – A Process Framework that Integrates Design Sprints, Agile Processes, and Machine Learning with Humans |date=2020 |editor-last=Degen |editor-first=Helmut |editor2-last=Reinerman-Jones |editor2-first=Lauren |series=Lecture Notes in Computer Science |volume=12217 |language=en |location=Cham |publisher=Springer International Publishing |pages=136–145 |doi=10.1007/978-3-030-50334-5_9 |arxiv=2003.05268 |isbn=978-3-030-50334-5|s2cid=212657395 }} The models also learn from human feedback and oversight. This amplification of learning on both sides is a key focus of RHML.

The approach draws on theories of learning in dyads from education and psychology. It also builds on human-computer interaction and human-centered design principles. Implementing RHML requires developing specialized tools and interfaces tailored to the application{{cite journal |last1=Zagalsky |first1=Alexey |last2=Te'eni |first2=Dov |last3=Yahav |first3=Inbal |last4=Schwartz |first4=David G. |last5=Silverman |first5=Gahl |last6=Cohen |first6=Daniel |last7=Mann |first7=Yossi |last8=Lewinsky |first8=Dafna |date=2021-10-18 |title=The Design of Reciprocal Learning Between Human and Artificial Intelligence |url=https://doi.org/10.1145/3479587 |journal=Proceedings of the ACM on Human-Computer Interaction |volume=5 |issue=CSCW2 |pages=443:1–443:36 |doi=10.1145/3479587|s2cid=239020698 |url-access=subscription }}

RHML has been explored across diverse domains including:

• Cybersecurity - Software to enable reciprocal learning between experts and AI models for social media threat detection.
• Organizational decision-making - RHML to structure collaboration between humans and AI systems.{{cite journal |last1=Shrestha |first1=Yash Raj |last2=Ben-Menahem |first2=Shiko M. |last3=von Krogh |first3=Georg |date=August 2019 |title=Organizational Decision-Making Structures in the Age of Artificial Intelligence |url=https://doi.org/10.1177/0008125619862257 |journal=California Management Review |volume=61 |issue=4 |pages=66–83 |doi=10.1177/0008125619862257|s2cid=199310764 |url-access=subscription }}
• Workplace training - Using RHML for workers to learn from AI technologies on the job.{{cite journal |last1=Grønsund |first1=Tor |last2=Aanestad |first2=Margunn |date=2020-06-01 |title=Augmenting the algorithm: Emerging human-in-the-loop work configurations |journal=The Journal of Strategic Information Systems |volume=29 |issue=2 |pages=101614 |doi=10.1016/j.jsis.2020.101614|s2cid=224771474 |doi-access=free |hdl=10852/78524 |hdl-access=free }}
• Open science - Using human and AI collaboration to promote open science.
• Production and logistics - turning workers and intelligent machines into teammates.

RHML maintains human oversight and control over AI systems, while enabling cutting-edge machine learning performance. This collaborative approach highlights the importance of keeping the human expert involved in the loop.

{{reflist}}

Category:Machine learning

Category:Human–computer interaction