In some patent applications, it is not clear whether the disclosed system produces a satisfactory, working, result. Or, the results can be very poor in quality terms (think software). Can an examiner determine an invention is useful, even if they cannot determine whether it really works, or works sensibly well enough?
An invention, as long as it is capable of being made or used in an industry, or, in other words has/satisfies “utility”, is considered useful, irrespective of whether the results delivered by it is poor in terms of quality. “Utility” or “Industrial Applicability” is one of the requirements of patentability. While reviewing a patent application, the Examiner looks at whether the disclosed subject matter, is capable of being made or used in an industry, irrespective of the quality of outcome. You can refer to the links below for more information on patentability requirements and non patentable subject matter under Indian Patent Act. http://www.invntree.com/blogs/what-can-be-patented-and-what-cannot-be-patented http://www.invntree.com/blogs/how-can-i-find-out-whether-my-invention-patentable
The US patent statute requires that an invention be "useful." 35 U.S.C. 101. Under the law, an inveention's utility must be "specific, substantial, and credible" and an invention lacking utility cannot be patented.
As a matter of practice, the utility doctrine is applied quite differently depending upon the context. For example, the US Patent Office requires evidence that a new pharmaceutical will work for its intended purpose rather than simply trusting the patent applicants claims that it works. On the other hand, the Patent Office's general practice has been to trust the applicant that the invention would actually work for its intended purpose.
The result then is that for software, the USPTO simply assumes utility in the vast majority of cases. I should say here that I expect that this will change over the next decade, but that is a long horizon.
It doesn't need to work optimally or even well. The "useful" criteria is used to reject perpetual motion machines and things that claim to cure all sicknesses. In medicine and biology they do want experiments you have carried out to show it does what you say it does. In mechanical or electronic fields it only has to be plausible - they do not have a lab to test anything out.