After establishing all your parameters and safeguards, take the survey questions or the poll, send it out with three different styles of the same data.  For instance, question #3 could become question #6 or rearrange all the components so that they are included but in differing forms.  Have the responses gathered and analyzed by two different means and follow with human analysis.  Compare all results and take the most obvious result.  Or, have the responders use two separate methods to answer; maybe one using number system and the other using alpha system, or electronically responding and compare that to human voice response to human recipient.