[Originally posted 6/10/2007]
Integrity in science means taking all necessary measures to ensure that neither conscious nor unconscious bias influences the result. This is why scientific methodology includes techniques such as double-blind studies and independent corroboration.
A recently released report in particle physics shows how "blind analysis" was used to avoid unconscious bias. The MiniBooNE experiment at Fermilab was designed to detect neutrino oscillations and confirm the anomalous results of a previous experiment at Los Alamos. The Jun 2007 issue of Physics Today reports that "to avoid unconscious bias, the group had estimated backgrounds and optimized all of its data-selection criteria without knowing how they would affect the final result. The experimenters got to "open the box" and look at what their data did...only three weeks before going public." The results failed to confirm the LANL result which had imposed rather awkward constraints on the standard model for neutrinos. But the data also revealed new puzzles that remain to be sorted out.
The point is that scientific methodology must always work to eliminate bias of any kind. The greatest care must be exercised when there is strong motivation to obtain a particular result. Those motivations may be confirming a particular theory, personal credit, company revenue or reputation, or religious preference.
In the long run, science is self-correcting and any bias-induced errors or fraud will be uncovered. The cost of uncovering such errors is great and needs to be avoided, particularly in large, expensive projects such as MiniBooNE. The principle holds for all scientific research, large and small.