The document discusses several issues with assessing the reliability of educational research. It notes that while some studies find positive correlations, such as practice tests improving performance, correlation does not prove causation. Confounding variables can also influence results if not properly controlled for. True experiments aim to minimize confounding factors by randomly assigning participants to control and treatment groups. However, some caution that educational experiments need to maintain real-world relevance. While student surveys often express positive attitudes, their actual behaviors may differ, so multiple measures are preferable to self-reports alone. The document also raises ethical considerations around only giving interventions to some students.