Observation and experiment: a causal perspective

Abstract

Observation and experiment and their related connotations and concepts remain vague, which affects the correct understanding of research design and the judgment of the validity of causal inference. This article borrows the concept of phase transition in physics, combines causal thinking and causal diagrams, firstly establishes the relationships among the attribute, state, event, and phenomenon, and then identifies two ways with the opposite causal structures to acquire phenomena-human observations and human manipulated experiments. In causal inference, the ways mentioned above, intervention and assignment of exposure are affected by their own causal mechanisms. Finally, intervention is a causal concept, a core link among known phenomena, unknown phenomena available for measurement, and natural causality. Based on this, the two strategies in classifying research design are analyzed, and intervention method and non-intervention method are proposed, as is comprehensive and concise. Observations and experiments provide the basis for all scientific knowledge and should be viewed as concepts with a unified connotation. The accurate classification of research designs based on the law of causality and measurement process may be one of the best options worthy of in-depth study.