Do you think AI is perfectly objective? Think again. A fascinating new study reveals how even the most advanced language models, like GPT-4, can be swayed by something as simple as word order. Just like humans, AI's judgment of similarity between things—say, comparing countries—shifts depending on which thing is mentioned first. This 'order effect,' a well-known quirk in human psychology, was tested on eight leading LLMs using pairs of countries. Surprisingly, several AI models displayed the same bias as humans, giving different similarity scores depending on the order of the countries in the question. For example, an AI might rate the similarity between North Korea and China higher than the similarity between China and North Korea, simply because of the order they were presented in. This isn't just a theoretical curiosity. As AI agents become increasingly integrated into our daily lives, from chatbots to decision-making systems, understanding these biases is crucial. Imagine an AI shopping assistant that suggests different products based on how you phrase your request. Or a news aggregator that prioritizes certain stories simply because of the order they appeared in a feed. The study’s findings highlight that even seemingly subtle contextual factors can significantly impact AI behavior. While sometimes this human-like bias can be helpful, in many scenarios it could lead to skewed or unpredictable outcomes. Researchers are now exploring why this happens, diving into the reasoning processes behind these AI judgments. Unraveling how context influences AI’s internal workings will be vital for building more reliable and robust AI systems in the future. So, the next time you interact with AI, remember: It's not entirely impartial. Just like us, it's shaped by the context surrounding it.