Page - 23 - in The Future of Software Quality Assurance

The Next Generation Tester: Meeting the Challenges ofa Changing IT World 23 • Detection of situations where too many matches are found and potentially incorrect responses are provided to requests (overfitting). This is an indication that matchingalgorithmsare toogeneralandprovideunwantedresponses. Challenges for testing • Any test needs expected results to enable detection of defects. Because AI applications are continuously “learning” and updating their knowledge bases, the definitionofexpectedresults isdifficult. • Testers must involve business users in the evaluation of results and in any decision to raise adefect report. Competencies requiredof theNG Tester: • Makestakeholdersawareof thekey risksofAI applications. • Coordinate and test specific aspects of AI applications in cooperation with business stakeholders. • Understand AI application architecturesand the specific purposeof the applica- tion. • Set up different sets of data in order to identify issues with underfitting and overfitting. 7 TheSkillsFramework So far we have considered the general competencies to be provided by the “next generation” tester and a selection of the specialist competencies. Whereas ideally all of the general competencies should be available, testers, employers and projects need to decide on which specialist competencies are required and where skills development should take place. Given the wide range of possible testing competencies, these decisions are best supported by a skills framework which goes well beyond the simple T-Model. A skills framework captures the following elements: • Competencydescriptions • Competency levels • A methodforassessing competencies • Profiles 7.1 CompetencyDescriptions Competency descriptions are provided of general and specialist competencies. At highlevel, thesewouldbesimilar to thoseoutlinedabove,butmaybefurtherbroken downas required.