Inductive inferences are based on observations and controlled experiments. To minimise bias and achieve a greater objectivity, a degree of personal detachment and independent verification is required. However, this should not lead to an objectivism that demands a researcher to be entirely neutral. Detachment, taken too far, can become an obstacle rather than an advantage. Positivistic science has conflated two different meanings of objective: ‘unbiased' and ‘external'. This is not only unnecessary, but also mistaken: perception of the external can be biased and of the internal can be impartial. By explicitly excluding the subject, reality can never be captured in its totality. Besides, scientific detachment can be only an ideal. Researchers cannot completely avoid bringing themselves into the story. So, distancing seems a more realistic attitude than detachment. While detachment strives to achieve objectivity by eliminating the subject, distancing does so by including and maintaining a larger perspective.
In any case, the value of empirical research cannot be denied, findings based on impartial observation and experimentation can greatly contribute to the understanding of physical reality. However, as already pointed out, it has been recognised that induction is not infallible, and should not be taken for granted. Moreover, some phenomena cannot (or at least not yet) be directly examined; they can only be deduced from their consequences. This is nothing new. For example, Silver writes that ‘molecules were part of the scientist's explanation of nature long before we could observe them. Their existence was deduced from the behaviour of matter' (1998, p.18). Through deduction we can arrive at certain knowledge that would not be accessible otherwise. Of course, it is important to go as far as possible in providing empirical support, but some conclusions will always have to be inferred.
There are several tools that can be used to assist this process, such as mathematics, geometry or theoretical (logical) conjectures. Nevertheless, even with the help of such systems, however stringently applied, deductive conclusions are not completely safe. In the 19th century non-Euclidian geometry was constructed, shortly before logic appeared to be not completely logical, and Gödel's theorem (mentioned above) showed that even mathematics is not foolproof. Furthermore, deduction cannot prove that its conclusions are true, because they depend on their premises that cannot be deduced.
To conclude, although this method can contribute to better understanding and finding more probable or plausible explanations, it should be recognised that both its components, induction and deduction, have their limits. This may not matter in some relatively simple cases, but more complete interpretations would again require the combination of inductive/deductive inferences with other methods.