Though many research-based problem types have been shown effective in promoting students' conceptual understanding and scientific abilities, the extent of their use in actual classrooms remains unclear. We interviewed and surveyed 16 physics and engineering faculty members at a large US Midwest research university to investigate how university instructors value and use different types of problems in teaching their introductory courses. Most of these instructors valued traditional textbook problems as a useful building block and used them frequently. They also valued non-traditional problems but seldom used them in their introductory courses. The divergence between value and use in part is due to limited resources and experience, time constraints, student reactions, and personal preferences. It is, however, also due to the fact that the beliefs held by some instructors regarding the nature of problem solving deviate from commonly accepted views in educational research. Although on average instructors were hesitant to reduce more than 20% of course content for research-based materials, they showed great interest in collaborating with educational researchers at a more personal level to increase the use of non-traditional problems in teaching. Our study indicates that instructors need not only research-based problems suitable for their introductory classes, but also assistance in how to effectively implement these problems.