JOURNAL OF
DECISION SYSTEMS
Generative AI, ChatGPT, and the Future of Human Decision Making

SUBMISSION DUE DATE:

March 31, 2024

GUEST EDITORS

Prof./Dr. Gloria Phillips-Wren, Loyola University Maryland, USA, gwren@loyola.edu
Prof./Dr. Ciara Heavin, University College Cork, Ireland, c.heavin@ucc.ie

CONTEXT AND RELEVANCE

ChatGPT, a generative pre-trained transformer, was released publicly on November 30, 2022, by OpenAI (https://openai.com/blog/chatgpt). It was described by the developers as “a sibling model to InstructGPT which is trained to follow an instruction in a prompt and provide a detailed response.” During the initial research phase, the developers said that they were “excited to introduce ChatGPT to get users’ feedback and learn about its strengths and weaknesses.” The “text-generating AI chatbot … is able to write essays, code and more given short text prompts, hyper-charging productivity. But it also has a more … nefarious side” (Stringer & Wiggers, 2023). Venture capital quickly flowed and continues to flow into this sector, and the technology arms race began between major technology companies that had been investing in this area for some time but were concerned about the potential for harmful effects (Swartz, 2023).

The technology is from a class of Artificial Intelligence methods (AI) referred to as ‘generative AI’ that have been investigated over the last ten or so years (Karpathy et al, 2016). Although generative AI is broadly applicable to other areas such as image and music generation, our focus in this special issue is its application to language and models relating to decision making, decision support, and decision support systems (DSS).

As a starting point, older neural networks (NN), also called artificial neural networks (ANN), are trained to recognize patterns and make decisions based on input data. NN gave been used routinely in DSS for applications such as fraud detection. One feature of NN is that they can modify or ‘learn’ as new data become available. They are designed in theory to mimic the way that a human brain processes data, and their applications expanded as computers achieved significant jumps in data storage and processing. A lingering criticism is the lack of transparency and explainability in the algorithms used to deliver the results, particularly when the system delivers a decision that affects people’s lives such as being denied access to a loan. Generative AI suffers from the same black-box opacity, but new algorithms are delivering seemingly expansive capabilities that have yet to be explored and understood.

Generative AI such as ChatGPT are referred to as Large Language Models (LLM) since they make use of massive amounts of textual data from sources such as web pages, databases, and program code. They hyper-jump the capabilities of NN and Natural Language Processing (NLP) by permitting a back-and-forth conversation between a human and the AI system that seems natural much of the time. To do so, the context of the conversation is needed to provide human-like responses during the human-to-bot conversation. LLM include a transformer as a type of NN to weight the significance of each part of the input data to provide context. Other specialized methods are used to tailor the conversation. Although ChatGPT can provide useful information as a starting point, it does not understand the meaning of the text that it generates, so it can be offensive, incorrect, nonsensical, biased, outdated, plagiarized, and possibly even dangerous.

This SI hopes to explore some of the nuances and use cases (Google, 2023; Scott, 2023) for the current state of this technology, particularly as we look to its possible impact on human decision making. For a recent panel discussion of ChatGPT and decision making, see https://www.youtube.com/watch?v=lvQR9H1sNvM.

OBJECTIVE OF THE SPECIAL ISSUE

The objective of this special issue is to investigate the impact of generative AI technologies such as ChatGPT on human decision making with the potential for both positive and negative consequences. High quality conceptual and empirical research papers are invited from the international interdisciplinary scientific community interested in decision making and decision support systems. We also are interested in practitioner viewpoints and actual use cases to spur debate and thought in this area.

JOURNAL OF DECISION SYSTEMS Generative AI, ChatGPT, and the Future of Human Decision Making

JOURNAL OF
DECISION SYSTEMS
Generative AI, ChatGPT, and the Future of Human Decision Making