[AI Ethics Seminar EP.3] Introducing Major AI Toolkits for Designing Trustworthy AI Experiences and Practical Application Guidelines - LG AI Research BLOG

With the advancement of generative AI, technical limitations such as Hallucination and data bias have created an environment where they can lead to corporate reputational and legal liability. Accordingly, the importance of designing safe and trustworthy Human-AI Interaction is growing.

Microsoft's HAX (Human-AI Experience Toolkit) is a framework used to design interactions between humans and AI. It provides a 4-stage interaction process and 18 guidelines to help AI understand user needs and earn trust, and in particular takes into account situations where the system does not work correctly.

Google's PAIR (People+AI Research) Guidebook focuses on transparency, explainability, and bias removal in AI models. It treats providing unique value to users as the measure of success, and offers a systematic guide covering everything from data collection to feedback and error management systems.

Based on its review of these global companies' Toolkits, LG AI Research has established and is utilizing its own AI Ethics Impact Assessment framework, which allows risk factors to be considered and system design to be concretized from the early stages of product development.