人工智能赋能的可信同伴互评模型构建与验证

Construction and Verification of Trusted Peer Evaluation Model Empowered by Artificial Intelligence

同伴互评是培养高阶思维、提升学习绩效的重要学习策略。然而在教学实践中,同伴互评存在严重的低信任问题,即学习者对同伴的评价存在质疑或漠视,整体接受度不高。究其缘由,学习者的认知差异性是导致低信任问题的关键诱因:客观上,学习者的认知差异性导致同伴评价的不一致现象;主观上,学习者面对评价不一致现象时会产生确认偏误。人工智能赋能的可信同伴互评模型,使用可信系数标明评价的有效性,引导学习者建立正确的认同与信任,可以降低学习者的确认偏误。该模型的智能化实现过程分为4个阶段:评价及可信系数的表征、评分关系加权图构建、基于加权随机游走算法的同伴间认知水平关系挖掘、可信系数计算及评价反馈。基于该模型的教学实践表明:其能够依据评价者与被评者的相对认知水平为评价计算合理的可信系数;有助于提高学习者对同伴评价的接受度,在感知有用性、行为意愿2个维度上显著优于传统同伴互评模型;对学习者批判性思维倾向的培养具有显著的正向影响,在分析性、系统性、求知欲和思想开放性4个子维度上均有显著提升。

Peer assessment is an important learning strategy to cultivate learners’high-order thinking and improve learning performance.However,in practice,peer evaluation has a serious problem of low trust,that is,learners question or ignore peer evaluation and the overall acceptance is not high.Cognitive differences among learners are the key contributing factor to this problem:objectively,cognitive differences among learners lead to inconsistent peer evaluations;subjectively,learners may experience confirmation bias when faced with inconsistent evaluations.The trusted peer evaluation model empowered by artificial intelligence uses a credibility coefficient to indicate the effectiveness of the evaluation to guide learners to establish correct identification and trust,which can reduce learners’confirmation bias.The intelligent implementation of this model can be divided into four stages:the representation of evaluation and credibility coefficient,the construction of weighted graph of scoring relationship,the mining of peer cognitive level relationship based on weighted random walk algorithm,the calculation of credibility coefficient and evaluation feedback.The teaching practice based on this model shows results as follows.First,the model can calculate a reasonable credibility coefficient for evaluation based on the relative cognitive level of the evaluator and the respondent.Second,learners’acceptance of peer evaluation has improved,especially in terms of perceived usefulness and behavioral willingness,which are significantly better than traditional peer evaluation models.Third,it has a significant positive impact on the cultivation of learners’critical thinking tendency,of which the four sub dimensions including analysis,systematicness,curiosity and openness have been greatly improved.