The New York Times devotes a long article to understanding why such a high fraction of students who begin science, technology, engineering and math (STEM) majors drop out of them in their second and third years of college.
The National Science Board, a public advisory body, warned in the mid-1980s that students were losing sight of why they wanted to be scientists and engineers in the first place. Research confirmed in the 1990s that students learn more by grappling with open-ended problems, like creating a computer game or designing an alternative energy system, than listening to lectures. While the National Science Foundation went on to finance pilot courses that employed interactive projects, when the money dried up, so did most of the courses. Lecture classes are far cheaper to produce, and top professors are focused on bringing in research grants, not teaching undergraduates.
I have an interesting experience with this problem. Getting students engaged with skeletal biology is really easy, because they can get started learning practical information really fast. This is a common pathway for students to enter biological anthropology. In genetics, in contrast, is has historically been harder to devise practical early experiences. First genetics courses are very much based on theory and memorization. Students who get onto a lab bench early are likely to stay very engaged. But many areas of biology today are most productively learned in other ways than the bench.
I've been putting undergraduate students directly into bioinformatics, getting them working with data and presenting theory as it becomes useful to the project. This has been a really positive experience so far, and there are just countless opportunities today to get students working with the sea of data coming from next-generation sequencing projects. But there's not really much support at hand for developing these practical experiences for students -- that's something that hasn't changed since the eighties. Very hard to envision scaling up to a broader set of undergraduates, because a lot of supervision is necessary for these experiences.
The article discusses piecemeal solutions, and more widespread ones adopted by some engineering schools for retaining first and second year students. Taking students who start out interested and engaged with science, and then treating the subject as "sink or swim", is a waste of everyone's time.
The irony is that everyone already treats lab experiences as the only serious training for STEM students in many fields. Professors already bring undergraduate students into labs and spend time (and their graduate students' and postdocs' time) training them in lab experiences. They just use the lecture classes as an expensive and time-consuming IQ test to filter those students. But this has a real cost: Instead of developing expertise within the undergraduates, which might get some real work done, and at least allow senior students to train younger cohorts, they learn techniques only a year or two before they depart.