LG AI Research's AI Ethics Seminar

While generative AI is maximizing productivity, concerns are being raised that it may simultaneously weaken humans' initiative in thinking. Recent research has shown a tendency for the effort to think for oneself to decrease as trust in AI increases.

Instead of blindly believing or doubting AI's outputs, users need to develop 'Appropriate Reliance'—accepting AI when it is right and defending their own judgment when it is wrong. To achieve this, the following strategies are needed when designing AI:

• Supporting Answer Verification: Presenting the basis for AI's answers so users can judge for themselves
• Uncertainty Notifications: Explicitly indicating that AI's answers may not be accurate, encouraging careful interpretation
• Thought-Provoking Questions: Instead of giving definitive answers, posing questions that ask for counterexamples to stimulate critical thinking

Ultimately, the core of AI design lies not simply in creating smart models, but in fulfilling the role of a partner that promotes human critical thinking.