Procedures for the assessment of risks or vulnerabilities from radiation in space are evaluated in terms of model-independent inductive approaches. The reliability of risks calculated for space applications on the basis of accelerator-based physical and biological data is examined from a microdosimetric perspective. Probability distributions for energy deposition in biologically significant sites extend over several decades in lineal energy even for monoenergetic high-energy particles of relatively high atomic number. Because the response depends on a large number of variables and because of the difficulty of incorporating all such factors into calculations, a precise correlation between a physical descriptor of the field and observed effects in space is not feasible. For the same reasons, it is equally difficult to estimate the accuracies of such risk assessments. We use recently published microdosimetric spectra for HZE particles and biological weighting functions, including those derived from biological measurements with maximum entropy techniques, to illustrate some problems associated with the evaluations of risks from radiation fields in space.