AIGC质量评估指标体系构建研究——以ChatGPT为例

Research on the Construction of Quality Evaluation Index System of AIGC*--Taking ChatGPT as an Example

[目的/意义]构建AIGC质量评估指标体系,有助于用户识别高质量的AIGC以及AIGC服务商提高内容质量。[方法/过程]通过文献检索和用户访谈获得原始数据,使用扎根理论研究方法对原始数据进行三级编码,提取AIGC质量评估指标。在此基础上,使用问卷调查法和层次分析法确定指标权重,构建AIGC质量评估指标体系。[结果/结论 ]AIGC质量评估指标体系包括内容特征、表达特征、效用特征、技术特征4个一级指标和个性化、高效性、易理解性、可解释性等21个二级指标,其中内容特征的权重最大,对于AIGC质量的影响程度最高。此外,研究发现AIGC与PGC、UGC在质量评估方面存在一定的评估异同,可追溯性、类人性、可塑性、可解释性等是AIGC质量评估特有的指标。

[Purpose/Significance]The construction of quality evaluation index system of AIGC is helpful for users to identify high-quality AIGC and AIGC service providers to improve content quality.[Method/Process]The original data were obtained through literature retrieval and user interviews,and were encoded at three levels based on grounded theory to extract the quality evaluation indexes of AIGC.On this basis,a questionnaire survey and an alytic hierarchy process were used to determine the weight of each index,and the quality evaluation index system of AIGC was constructed.[Result/Conclusion]The quality evaluation index system of AIGC includes 4 first-level indexes such as content characteristics,expression characteristics,utility characteristics and technology charac teristics,along with 21 second-level indexes such as personalization,efficiency,intelligibility and interpretability.Among these,content characteristics have the highest weight and the greatest impact on AIGC quality.In addition,the study finds that AIGC,PGC and UGC have some similarities and differences in the evaluation of information quality.Traceability,humanism,plasticity and interpretability are unique indexes of AIGC quality evaluation.