According focused because in some projects 50%
According to Salman (2016), software quality and testing are insufficientregarding cognitive biases.
Softwarequality is affected when human thinking process differs from the laws, and that refers to cognitive biases. The author discusses different factors of cognitive biasesin an organization and how software quality and testing is affected. Softwarequality can be improved by focusing oncognitive biases and the factors thatgenerate them (Salman, 2016).The software development lifecycle (SDLC) consists ofdifferent phases where humans perform the tasksin multiple roles. As the quality andtesting engineers, work done in the testingphase to improve the quality includes technology, processes along with human thoughts.
The impact of cognitivebiases in the quality and testing phaseshould be focused because in someprojects 50% of the overall project cost depends on this phase (Salman, 2016).According to Salman (2016), many studies have been conducted, anddifferent cognitive biases such asconfirmation bias, representative’s, and anchoring were focused. In (2016),Salman performed a study on different cognitive biases and how variousfactors generate biases in an organization software testing process. The factors in an organization include company culture, team structure, timepressure (Salman, 2016). The study analyzed various effects of biases onsoftware testing and proved that software quality is affected by using afunctional test strategy.
Also, high production defects occur because of confirmationbias among software testers.In (2016), Salman was performinga systematic literature review and experimental case study at the time of the article’spublication. The literature review focused on the area of software engineering andidentified different cognitive biases factors that generate biases.
Informationfrom the literature review is taken as the input to conduct the case study. A time-pressure experimental case study is considered because time pressure exists in the softwareindustry. The case study discovers if time pressure generates confirmation biasin software testers during testing.According to Salman (2016), onlycognitive confirmation bias was observed, and the other cognitive biases are tobe explored and validated yet. In (2017) Knauss et al. discusssoftware-related challenges of testing automatedvehicles by conducting a case study and gathering information from two groupssuch as focus groups and interviews. The survey includes participants such asvehicle manufacturers, suppliers, andresearchers from 5 different countries.
The advancements in software technologypermitted vehicles to support conditionally automatic driving. While thisfunctionality has advantages but safety is a concerned issue as the studyindicates many vehicles involving inaccidents.According to Knauss et al.(2017), testing, verification, validation, andcertification are some of the primarychallenges faced by the automated road vehicles. The author discusses some ofthe testing difficulties associated withsoftware engineering that were identified from the case study.
Firstly, virtual testing and simulation is an approach that focuses on theautomation levels. In this approach,virtual testing is considered as the practical testing is expensive. The challengesfaced by the virtual reality testing arelegal aspects, mixed reality testing, integration of techniques with thetesting process, and gathering real traffic data for simulation testing.
The second is a changeof responsibility where the driver of thevehicle shifts the functionality to automatic mode. The functions of the vehicleshould be robust as it is in charge for eachpart of driving. Also, the software related non-functional requirements such asuser interface must also be tested from the driver’s aspect.
The third isdefining standards, regulations, and certifications for the automated vehicles.The fourth is automatictesting in which ability of testing increases with low costs and few steps. Forexample in a fully automated vehicle testing must be a continuous process wherethe testing process is based on specific criteria. The fifth challenge is selectingthe development process.
The requirements of the automated vehicles are not entirelyavailable during the design phase. The author cites that integrating V-modeland continuous software engineering process will be helpful.Finally, testing of artificialintelligence(AI), testing of thefunctionality that is implemented involves two properties such as self-adaptiveand artificial intelligence with sample data. During the run-time, the vehiclemust improve the ability to test automated systems. The current processes donot test the techniques of machine learning during runtime, it requires new AIand optimization tools.
