Impacto de la Inteligencia Artificial Generativa en la Comunicación Científica

The integration of knowledge generated by generative artificial intelligence models has shown a significant impact on the efficiency of various productive processes but raises challenges related to misinformation, privacy, security, and biases. In scientific communication, these models face the issue of hallucinations—incorrect or fabricated content—that can be mitigated by configuring the "temperature" variable in prompt design. This study, using a multimethod approach that includes narrative review, quasi-experimental experiments, and methodological triangulation, analyzes how this configuration affects the quality and accuracy of the generated text. The findings reveal that low temperature values (close to 0) enhance precision and reduce hallucinations, while high values increase creativity but also heighten the risk of speculative and inaccurate responses, as well as extend text length. The study concludes that temperature configuration should be tailored to the specific objectives of scientific communication, prioritizing precision in contexts where accuracy is critical. From an ethical perspective, improper use of this variable can lead to misinformation if results are not rigorously verified and validated. The study offers practical recommendations for balancing creativity and precision, emphasizing that careful temperature adjustment not only mitigates risks but also optimizes the generation of reliable content, thereby strengthening scientific communication.