While I won't dispute assertions like Abdhulrahman's that trial and error is best “used if you are in the process of learning or not in an online machine (which is in production)” (Alrahbi, 2010) I must note that we are always learning and any action whatsoever can have unforeseen consequences. So when Mohamed says “It's not recommended to use trail and error because it's tedious and time consuming. The better is to plan and work your plan” (Eraky 2010) I feel this is a sweeping generalization. Trial and error can be tedious and time consuming but this is not a given. And plans can be destroyed by unforeseeable events no matter how detailed or full of contingencies. If one has enough information to make an 'educated guess' what one even means by trial and error is a bit amorphous.
As David says, “Resorting to 'Trial and Error' to resolve a problem usually indicates to me that I have not (or maybe cannot) fully understood the principle of the problem” (Cook, 2010). Not fully understood the principle. In all honesty there is precious little in this world of which I can say I fully understand the principle. I am forced to make an awful lot of assumptions if I want to get anything done. I need to accept as axiomatic theories, and yes even hypotheses that I am unqualified to analyze nor for which do I have any first-hand empirical evidence. So just about everything I do can be said to be trial and error to some degree.
Mohamed also makes an apparently valid point about Kevin's “Analogy, where the algorithm under development could be compared with a previously developed and working sample, in order to come up with a working solution.” (Ouma, 2010) when he says “This step will not be effective if the problem is newest and is not happened before and no experience about it” (Eraky, 2010). But I would suggest that we unpack this argument a little. Not effective? Analogies are not inherently functional, they're merely an aspect of how we think (Holyoak et al, 2001). If the problem is indeed newest and we have no experience, analogizing to any experience we do have may be the only grasp we can get on it. I think Mohamed is probably correct and most of our attempts will be useless and even potentially damaging but that won't stop us from making them. And there will even be times when we get 'lucky'.
So, I submit to you that trial and error is our strength. If and when machines can err as effectively as we we will have truly created peers (Whitehead and Ballard, 2010). In closing I'll simply admit that on my own systems I am utterly a loose cannon (yet another military analogy), I don't care if I break it because I will learn something. For clients, on the other hand, I suggest a three-tiered (production, qa and development) architecture at least.
Alrahbi, Abdhulrahman (2010) RE: DQ1: Problem solving techniques [Online]. Available from: https://elearning.uol.ohecampus.com/webapps/portal/frameset.jsp (Accessed: 17 February, 2010).
Cook, David (2010) RE: DQ1: Problem solving techniques [Online]. Available from: https://elearning.uol.ohecampus.com/webapps/portal/frameset.jsp (Accessed: 17 February, 2010).
Eraky, Mohamed (2010) RE: DQ1: Problem solving techniques [Online]. Available from: https://elearning.uol.ohecampus.com/webapps/portal/frameset.jsp (Accessed: 17 February, 2010).
Holyoak, K. J., Gentner, D. and Kokinov, B.N (2001) “The Place of Analogy in Cognition”, Perspectives from cognitive science. Cambridge, MA: MIT Press [Online]. Available from: http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.72.1467&rep=rep1&type=pdf (Accessed: 17 February, 2010).
Ouma, Kevin (2010) RE: DQ1: Problem solving techniques [Online]. Available from: https://elearning.uol.ohecampus.com/webapps/portal/frameset.jsp (Accessed: 17 February, 2010).
Whitehead, S.D. And Ballard, D.H. (1991) “Learning to perceive and act by trial and error”, Machine Learning, Volume 7, Number 1, July, p45-83 [Online]. Available from: http://www.springerlink.com/content/j1435551574j5685/fulltext.pdf (Accessed: 17 February, 2010).