Source-linked AI summary
Generative Artificial Intelligence: Evolving Technology, Growing Societal Impact, and Opportunities for Information Systems Research
Veda C. Storey, Wei Thoo Yue, J. Leon Zhao, Roman Lukyanenko
TL;DR
The paper asks how GenAI’s distinctive features and impacts should be understood and how information systems research should respond. It traces AI’s evolution and develops a sociotechnical systems framework to connect technical and organizational perspectives. The resulting agenda treats GenAI as a generative sociotechnical system while recognizing limits in its human-like understanding and risks including intellectual-property infringement.
Problem
It remains unclear how GenAI’s distinctive features create transformative potential and what perspective IS research can provide on its impacts on individuals and organizations.
Method
The paper retraces AI’s evolution and develops a systems-theory-based sociotechnical framework for understanding GenAI and guiding an IS research agenda.
Results
The paper proposes GenAI as a generative sociotechnical system and identifies research topics concerning its business and societal impacts.
Takeaways & Limitations
IS research should examine GenAI as part of broader sociotechnical systems, including its organizational integration, human-AI relations, and unexpected outcomes.
Takeaways & Limitations
LLMs can produce coherent, human-like text but do not understand language in the human sense or reliably distinguish factual from nonfactual information.
Abstract
from arXiv · showhide
The continuing, explosive developments in generative artificial intelligence (GenAI), built on large language models and related algorithms, has led to much excitement and speculation about the potential impact of this new technology. Claims include AI being poised to revolutionize business and society and dramatically change personal life. However, it remains unclear exactly how this technology, with its significantly distinct features from past AI technologies, has transformative potential. Nor is it clear how researchers in information systems (IS) should respond. In this paper, we consider the evolving and emerging trends of AI in order to examine its present and predict its future impacts. Many existing papers on GenAI are either too technical for most IS researchers or lack the depth needed to appreciate the potential impacts of GenAI. We, therefore, attempt to bridge the technical and organizational communities of GenAI from a system-oriented sociotechnical perspective. Specifically, we explore the unique features of GenAI, which are rooted in the continued change from symbolism to connectionism, and the deep systemic and inherent properties of human-AI ecosystems. We retrace the evolution of AI that proceeded the level of adoption, adaption, and use found today, in order to propose future research on various impacts of GenAI in both business and society within the context of information systems research. Our efforts are intended to contribute to the creation of a well-structured research agenda in the IS community to support innovative strategies and operations enabled by this new wave of AI.
1 Introduction
GenAI has intensified speculation about AI’s transformative effects while leaving open how society and information systems research should respond. The paper addresses this gap by framing GenAI as a generative sociotechnical system and proposing an IS research agenda.
- GenAI’s emergence: GenAI systems produce text, images, music, programming code, and other complex creative outputs, with LLMs enabling human-language generation from inputs.Recent advances have made these systems capable of processing large amounts of data and performing knowledge-centric tasks.
- GenAI’s emergence: The rapid popularity of tools such as ChatGPT and Dall-E has generated substantial interest and speculation about GenAI’s everyday and organizational roles.Analyses range from broad productivity applications to concerns about plagiarism and misinformation.
- Research gap: GenAI raises unanswered questions about how it transforms business activities and how to ensure productive, ethical, safe, and responsible use.These questions motivate the paper’s central inquiry into the perspective information systems research can provide on GenAI’s impacts.
- Paper’s response: The paper develops a theoretical framework that uses systems theory and a sociotechnical lens to examine GenAI’s components, behavior, and organizational integration.The framework also draws on foundations such as linguistic theory and supports a research agenda for IS scholarship.
- Paper’s response: The paper proposes the notion of a generative sociotechnical system and identifies future research topics for understanding GenAI through existing IS concepts and theories.It also argues that GenAI’s capacity to generate unexpected results requires boundary conditions and new approaches to research.
2 Related Research
This section situates GenAI within prior AI and IS research by tracing major shifts from foundational and symbolic approaches to machine learning, deep learning, transformers, and LLMs. It shows how earlier AI advances and IS engagement provide the basis for studying GenAI’s current implications.
- Prior IS research: IS research has examined how successive information technologies, including data processing, analytics, enterprise systems, and electronic commerce, reshape business and social practices.This history positions IS as a field that studies technology adoption and organizational use.
- Prior IS research: Figure 1 maps major AI innovation stages above the line and related IS research areas below it, showing IS research’s continuing engagement with AI’s progression.The figure provides the historical foundation for studying GenAI and its applications from an IS perspective.
- AI evolution: Big data, expanding computing power, and new algorithms accelerated machine learning and deep learning, with ImageNet marking a major 2009 breakthrough.These developments changed research and development across business, science, and engineering domains.
- From symbolic AI to GenAI: The 2017 transformer architecture enabled large language models through self-supervised learning and self-attention over large internet text corpora.Transformer-based models rapidly approached or surpassed human-level benchmarks on GLUE and SuperGLUE evaluations.
- From symbolic AI to GenAI: The current AI breakthrough centers on LLMs and associated algorithms, reflecting a shift from symbolism toward connectionism and natural-language generative applications.Although post-2017 systems emphasize language and generation, their concepts and algorithms evolved from earlier AI efforts.
3 Theoretical Framework for Understanding Generative AI
The framework presents GenAI as a complex sociotechnical system whose generative properties distinguish it from traditional AI and create new organizational and societal research opportunities. It emphasizes strong emergence, novelty, systemic inputs and outputs, and the need to study GenAI within broader human, organizational, and technological systems.
- Conceptualization of Generative AI: GenAI is conceptualized as an advanced question-and-answer system that transforms large-scale data into human-like responses to natural-language prompts.Its computational foundations include transformers, attention mechanisms, GANs, reward models, and RLHF.
- Traditional versus Generative AI: Unlike conventional Q&A systems that retrieve prestored keyword-matched answers, GenAI generates novel, complex, and self-contained outputs.Traditional AI generally focuses on bounded decision spaces, whereas GenAI transforms data into new connections and outputs.
- Generative Properties: GenAI produces potentially unlimited outputs from similar inputs, with novelty arising through new transformations and connections among very large parameter and data spaces.The resulting behaviors are difficult to predict exhaustively, especially as systems combine multiple data types.
- Generative Properties: GenAI can accept and generate coherent component systems as well as standalone systems, including essays, scripts, images, animations, music, and responses.These outputs can be used directly or embedded within broader organizational technologies such as decision-support systems.
- Generative Properties: GenAI exhibits strong emergence because outputs arise from prompt transformations combined with complex system knowledge and are not directly derivable from individual components.This can support creative content while complicating control and assurance against harmful or disadvantageous outcomes.
- Sociotechnical Perspective: The sociotechnical framework places GenAI’s technical properties within systems, organizations, individuals, and processes to guide research on design, use, and impact.The proposed agenda includes human-AI collaboration, productivity, innovation, adoption, trust, job displacement, retraining, and the technology’s potential threats.
4 Research Opportunities for Information Systems Research
The paper develops a sociotechnical research agenda for examining GenAI’s effects on organizations and society, spanning human-AI collaboration, system design, value, and risks. It highlights opportunities to study GenAI’s transformative capabilities alongside constraints involving opacity, unintended consequences, intellectual property, misinformation, manipulation, bias, and environmental impact.
- Research agenda: A sociotechnical research agenda should examine GenAI at the intersection of technical properties and individuals, organizations, and society.The proposed themes build on sociotechnical foundations and prioritize relationships between GenAI’s technical properties and social entities.
- Business value: GenAI research should assess business value, including its effects on efficiency, decision-making, productivity, and creative work.The paper calls for evaluating GenAI’s value within broader business contexts rather than treating technical capability as sufficient evidence of value.
- Human-AI collaboration: Human-AI collaboration research should define responsibilities, delegation, performance evaluation, interfaces, and configurations in which AI agents increasingly replace or support human work.The agenda includes human agency, prompt-based interaction, human-AI configuration, and design principles for AI-intensive systems.
- Human-AI collaboration: GenAI’s open-ended interaction and opaque LLM formation create research needs around prompt engineering, contextual communication, and managing fluid system-human exchanges.Users may need more specific contextual information for accurate responses, while not all users know how to formulate effective prompts.
- Risks and safeguards: GenAI’s adaptability raises risks involving unintended consequences, intellectual-property infringement, misinformation, deepfakes, emotional manipulation, bias, and hallucinations.The paper connects these risks to the need for oversight, governance, ethical safeguards, and research on responsible use.
- Environmental impact: GenAI’s resource demands motivate research on algorithms that minimize energy use and on the immediate and delayed environmental effects of widespread adoption.Training ChatGPT-3 reportedly required approximately 936 MWh, illustrating the scale of the energy issue discussed.
5 Conclusion
The paper frames GenAI as a widely accessible technology with potential to affect more business operations than previous technologies, while prompting new questions about its role and impacts.
- GenAI’s ready availability and natural-language interface enable diffusion without extensive user training.
- The paper examines GenAI as the next generation of AI in response to its rapidly expanding applications.
Declarations
The declarations state that the research is academic and unaffiliated with commercial companies or agencies, and that the article is openly licensed under Creative Commons Attribution 4.0.
- The authors declare that the work is pure academic research and is not related to any commercial company or agency.
- The article permits use, sharing, adaptation, and reproduction under a Creative Commons Attribution 4.0 license with required attribution and change notices.