Métricas y guías para el desarrollo de algoritmos de Inteligencia Artificial Explicable
Fecha
2021-09-06
Autores
Título de la revista
ISSN de la revista
Título del volumen
Editor
Jaén: Universidad de Jaén
Resumen
Una de las principales dificultades a la hora de desarrollar modelos de inteligencia artificial explicable, es que no
existe una definición única, común y aceptada en la bibliografía específica sobre algunos conceptos y métricas
que faciliten cuantificar la interpretabilidad de un modelo. En este trabajo se ha realizado una revisión de la
terminología relacionada con inteligencia artificial explicable en la bibliografía especializada, que permite
aportar definiciones integradoras, y proponer la unificación y agrupación de conceptos. Es un análisis inicial en
un área que debe afrontar muchos retos, entre ellos el debate, consenso y propuesta sobre conceptos y
métricas objetivas que cuantifiquen la explicabilidad de un modelo.
One of the main difficulties when developing models of explainable artificial intelligence is that there is no single, common and accepted definition in the specific literature of some concepts and metrics that facilitate quantifying the interpretability of a model. In this work, a review of the terminology related to explainable artificial intelligence in the specialized bibliography has been carried out, which allows providing integrative definitions, and proposing the unification and grouping of concepts. It is an initial analysis in an area that must face many challenges, among them the debate, consensus and proposal on concepts and objective metrics that quantify the explainability of a model.
One of the main difficulties when developing models of explainable artificial intelligence is that there is no single, common and accepted definition in the specific literature of some concepts and metrics that facilitate quantifying the interpretability of a model. In this work, a review of the terminology related to explainable artificial intelligence in the specialized bibliography has been carried out, which allows providing integrative definitions, and proposing the unification and grouping of concepts. It is an initial analysis in an area that must face many challenges, among them the debate, consensus and proposal on concepts and objective metrics that quantify the explainability of a model.