생성형 인공지능과 ADR —대체적분쟁해결을 위한 생성형 인공지능 활용에 대한 소고—

ChatGPT has ushered in a revolutionary new era. The vast knowledge of artificial intelligence (AI) and its ability to engage in natural conversations based on this expertise are expanding the horizons of technological advance- ment. With its vast knowledge and ability to engage in natural conversations, it has challenged and expanded the horizons of new technology. For example, generative AI models such as ChatGPT present opportunities to complement or enhance the traditional methods of alternative dispute resolution (ADR). This paper defines “one of the initial ADR systems utilizing AI” as “AIDR”, with a particular focus on the use of ChatGPT. The following sections will explore, among other topics, the applications, guidelines, and legal discussions surrounding AIDR. ChatGPT, as a generative AI, represents the latest evolution of chatbot technology. In 2018, OpenAI introduced the GPT language model, which garnered significant attention in the field of natural language processing, and then followed it with the release of GPT-4 in 2023. Meanwhile, AI has been used in ADR since the 1970s. Early AI models were rule-based systems in which experts manually crafted rules, a method that had its limitations. Despite these challenges, rule-based systems laid the foundation for large-scale consumer e-commerce platforms such as eBay and PayPal. The rise of machine learning and big data further transformed AIDR, allowing for more sophisticated functionalities that can generate solutions independently. Consequently, AIDR has become an area of increasing interest, especially with the advent of generative AI. AIDR systems can be categorized into supportive technologies and auto- mated technologies. Current AIDR applications lie somewhere between these two stages, with varying degrees of automation and human involvement. Upon reviewing the discussions on the introduction of AIDR, it can be observed that there is some hesitancy in adopting it, primarily due to the technology’s inability to fully explain its reasoning, raising concerns that relying on such systems may be premature. However, many countries like the U.S., as well as European ones, are advocating for introducing AI into the ADR market. Legal services based on GPT-4, such as CoCounsel, launched in March 2023 by the U.S.-based legal tech company Casetext, have already emerged. This service’s AIDR applications include preventing disputes, selecting neutral third parties, improving procedural efficiency, drafting preliminary rulings, and facilitating arbitration and negotiation. It is important to note that, to ensure the effective and appropriate use of AIDR, operational guidelines are essential. An example of regulatory guidance can be found in the policies of the American Arbitration Association-International Centre for Dispute Resolution (AAA-ICDR), reported in November 2023. These guidelines emphasize that AIDR must be supervised by human experts and deployed responsibly, with a clear understanding of its advantages and limitations. They also highlight the need for special care when handling sensitive data and recommend continuous improvements to the AI tools employed in ADR processes. One critical issue in applying AIDR is the hallucination phenomenon, which persists even in ChatGPT’s latest version, GPT-4. This refers to the tendency of AI to generate plausible-sounding but inaccurate or misleading responses. Additionally, bias remains a significant concern. For example, in actual tests, AI-generated images depicted the founder of a company as a white male, a secretary as a white female, and a domestic worker as a woman of color. These biases pose potential risks when AI is applied in ADR processes. Another crucial aspect of AIDR is data privacy protection. Many AI tools operate on cloud-based platforms, requiring the transmission of dispute-related data to third-party cloud service providers. This raises the risk of personal information leakage. Furthermore, advances in AI technology have enabled the creation of deepfakes, which could lead to manipulated or falsified evidence being submitted during ADR proceedings. If such false evidence is introduced, it can create confusion in determining facts, significantly undermining the reliability and fairness of the process. Concern also exists about delegating decision-making functions to AI without informing the parties involved. Failing to disclose such delegation to dispute parties could violate due process. In serious cases, it could provide grounds for invalidating or canceling the outcome of the ADR process. Furthermore, discussions are ongoing about whether AI could move beyond an assistive role to serve as a fully neutral third party with final decision-making authority. Some argue that since AI can analyze vast amounts of data, its decisions might be less biased and more accurate than those made by humans. However, a counterpoint is that AI lacks emotional intelligence and struggles to recognize subtle nuances, which could hinder its ability to produce fair and practical outcomes. It is likely that some parties opting for AI in dispute resolution will factor these considerations into their decisions. Moreover, with the recent advancements in AI’s ability to detect emotions, AI could potentially expand its role as a neutral third party in certain fields, such as when AI detects the emotional escalation of disputing parties and de-escalates the conflict situation. Additionally, in jurisdictions where there is no explicit requirement that a neutral third party must be a natural person, the eligibility of AI as a neutral party could be interpreted more openly. In summary, ADR has gained attention for offering simpler, faster, and more cost-effective procedures compared to litigation. It also provides flexible processes that encourage parties to reach agreements and explore solutions autonomously. However, in some cases, ADR can be just as complex as litigation, and high costs may arise depending on the expertise of the neutral third party, potentially undermining its intended benefits. Overall, the introduction of AIDR is expected to reduce time and costs while delivering more objective and fair decisions. To ensure the effective use of AIDR, it is essential to establish regulatory guidelines and engage in legal discussions regarding potential challenges. Based on the outcomes of these discussions, AIDR is likely to play an increasingly important role in enhancing the efficiency and trustworthiness of dispute resolution processes. In short, “AI is not a competitor to humans but a collaborator.” It is expected that the use of AIDR will continue to expand actively as a viable dispute resolution method in the future.