Adverse effects of AI: risks in the information age, disinformation and machine learning chatbots with explainable AI (XIA) approach: from A.L.I.C.E. to DeepSeek

Abstract. The paper analyses the adverse effects of artificial intelligence, particularly in chatbots, focusing on the risks of misinformation, polarization and algorithmic discrimination. The evolution of chatbots is explored, from rule-based systems to advanced language models such as A.L.I.C.E.®, Replika®, ChatGPT®, Bard® and DeepSeek®. This entails the hypothesis that the increasing sophistication of chatbots, from rule-based systems to advanced language models, leads to a rise in the risks of misinformation, polarization, and algorithmic discrimination, thus, Explainable Artificial Intelligence (XIA) emerges as a crucial tool to mitigate these adverse effects, although its application in different chatbots presents strengths and weaknesses in terms of explainability. The methodology used combines the responsible and ethical use of AI, driven by XIA, that essential to ensure that technology benefits society. Furthermore, Explained Artificial Intelligence (XIA) is proposed as a solution to mitigate these risks through explainability of AI systems. The results suggest a relationship in the application of XIA principles in different chatbots is evaluated, identifying their strengths and weaknesses in terms of explainability principles. The paper concludes by highlighting the importance of XAI for a responsible and ethical use of AI.