Page - 80 - in The Future of Software Quality Assurance

80 R. Marselis machine intelligence is way beyond the impact that previous technologies had on our businesses and our society. Testing of machine intelligence is different from traditional testing of information processing. With traditional IT, the tester could alwayspredicttheoutcomeofatestsincetherulesweredefinedupfront.Incomplex systems, it maybe a tough task but fundamentally it’s alwayspossible todefine the expectedoutcome. But now we have learning machines. Based on the input they gather using all sortsof sensors, theypickup informationfromtheir environmentandbasedon that determinethebestpossibleresultat thatgivenpoint in time.Bydefinition, theresult at another point in time will be different. That’s a challenge during test design and testexecution.Forexample,thisrequiresthat testersworkwithtolerancesandupper and lowerboundariesforacceptable results. Business Impact Newtechnologieshavealwayshadimpactonbusinesses.But theveryfastevolution and the impressive possibilities of the implementation of intelligent machines in businessprocesses requirespecial attention for thebusiness impact. As quality-minded people, we want to make sure that machine intelligence positively contributes to business results. To ensure this, IT teams need to use bothwell-knownandbrand-newqualitycharacteristics toevaluatewhether thenew technologycontributes to thebusinessvalue. Social Impact Until now, machines always supported people by extending or replacing muscle power. Now we see machines that replace or extend human brain power. This may have tremendous consequences for the way people interact and for society in a broader sense, which brings us to the last angle of quality: social impact. New technologieshave had social impact. So, what’s new this time? It’s the speed with which this technologyentersour lives. And the possibility of machines taking decisionsforpeople,eventotheextentwherepeopledon’tunderstandthereasoning behind thechoice fora specific option. 2.2 Quality Is DefinedandTestedUsingQuality Characteristics To get a clear view on the quality level of anysystem, we need to distinguishsome subdivision of quality. Therefore, we use quality characteristics. The long-known standards evolved in an era when IT systems were focused on data processing and where input and output were done by means of files or screen-user-interfaces. We use the ISO25010 standard as a basis for our elaboration. In Figure 2 you see the traditional ISO25010 quality characteristics in gray. In red you see the quality characteristics for intelligentmachines thatwe haveadded(Fig.2).