大数据分析中的相关性和因果关系

Correlation and Causality in Big Data Analysis

“大数据分析不追求因果关系而只关注相关性”是一种颇为流行但似是而非的说法。实际上,大数据分析并非完全放弃对因果关系的追求,其所关注的相关性是对因果关系的逼近和靠拢,是在无法确定因果关系时的一种折中,这与法律上的因果关系在大多数情况下属于统计的因果关系(强相关)并行不悖。大数据分析的结果可以在法律程序中适用,但受制于数据质量、建模错误等因素,其可靠性有时比较薄弱,从而导致其适用范围存在限制。大数据在法律程序中的作用主要是预警和佐证,仅在少数情况下才可以直接据以作出法律决定。大数据技术具有两面性,在充分利用大数据带来的便利的同时,也需要在观念上破除“大数据的神话”,特别是对大数据的伪相关性风险进行防范,避免可能的“大数据的悲剧”。

The saying that“Big data analysis does not pursue causality but only focuses on correlation”is a popular but specious statement.In fact,big data analysis does not completely abandon the pursuit of causality,its concern on correlation is the approximation and proximity of causality.That is a compromise when the causality cannot be determined,and that means it is not contradictory with the legal causality that belongs to the statistical causality(strong correlation)in most cases.The results of big data analysis can be applied in legal procedures,but due to the influence of data quality,modeling errors and other factors,its reliability is sometimes relatively weak,thus leading to limitations in its scope of application.The role of big data in legal procedures is mainly“early warning”and“backing”.Only in a few cases,legal decisions can be made directly based on big data.Owing to big data technology is double-edged,while making full use of the convenience brought by big data,it is also necessary to dispel“the myth of big data”in concept,especially to prevent the spurious correlation risk,and to avoid the possible“tragedy of big data”.