Artificial-Intelligence-and-Software-Engineering-S

ArtificialIntelligenceandSoftwareEngineering:StatusandFutureTrendsJörgRech,Klaus-DieterAlthoffThedisciplinesofArtificialIntelligenceandSoftwareEngineeringhavemanycommonalities.Bothdealwithmodelingrealworldobjectsfromtherealworldlikebusinessprocesses,expertknowledge,orprocessmodels.Thisarticlegivesashortoverviewaboutthesedisciplinesanddescribessomecurrentresearchtopicsagainstthebackgroundofcom-monpointsofcontact.1IntroductionDuringthelastdecadesthedisciplinesofArtificialIntelli-gence(AI)andSoftwareEngineering(SE)havedevelopedseparatelywithoutmuchexchangeofresearchresults.InAIweresearchedtechniquesforthecomputationsthatmadeitpossibletoperceive,reason,andact.ResearchinSEwasconcernedwithsupportinghumanbeingstodevelopbettersoftwarefaster.Today,severalresearchdirectionsofbothdisciplinescomeclosertogetherandarebeginningtobuildnewre-searchareas.SoftwareAgentsplayanimportantroleasresearchobjectsinDistributedAI(DAI)aswellasinagent-orientedsoftwareengineering(AOSE).Knowledge-BasedSystems(KBS)arebeinginvestigatedforlearningsoftwareorganizations(LSO)aswellasknowledgeengineering.AmbientIntelligence(AmI)isanewresearchareafordis-tributed,non-intrusive,andintelligentsoftwaresystemsbothfromthedirectionofhowtobuildthesesystemsaswellashowtodesignthecollaborationbetweenambientsystems.Lastbutnotleast,ComputationalIntelligence(CI)playsanimportantroleinresearchaboutsoftwareanalysisorprojectmanagementaswellasknowledgediscoveryindatabasesormachinelearning.Furthermore,inthelastfivetotenyearsseveralbooks,journals,andconferenceshavefocusedontheintersectionbetweenAIandSE.Theinternationalconferenceandasso-ciatedjournalAutomatedSoftwareEngineering(ASE)pre-sentsresearchaboutformalandautonomicapproachestosupportSE[2].SimilartopicswithastrongerfocusonKBSandknowledgemanagementarepublishedintheinterna-tionalconferenceandassociatedjournalofSoftwareEngi-neeringandKnowledgeEngineering(IJSEKE)[1].Inthispaper,wegiveashortoverviewaboutthestatusandfuturetrendsintheintersectionbetweenAIandSE.Wefocusonthetopicssoftwareagents,KBS,AmI,andCIastheareascoveredbythecontributionsofthisspecialissue.InSection2wedescribethedisciplinesAIandSE.ThefocusedtopicsaredescribedinmoredetailinSection3.Finally,inSection4wegiveanoutlookforthenextyearsandpresentnewchallengesforbothdisciplines.2ArtificialIntelligenceandSoftwareEn-gineeringThissectionwillshedsomelightonthedisciplinesAIandSEforthosenotfamiliarwiththeotherdiscipline.AspectsofArtificialIntelligenceThereisageneralagreementintheAIcommunitythatthedisciplineofAIwasbornattheDartmouthconferencein1956.AccordingtoWinston[81]“AIisthestudyofthecom-putationsthatmakeitpossibletoperceive,reason,andact”.Wachsmuth[78]assumesthisdefinitionandpointsoutthat,“AIdiffersfrommostofpsychologybecauseofitsgreateremphasisoncomputation,anditdiffersfrommostofcom-putersciencebecauseofitsgreateremphasisonpercep-tion,reasoning,andaction”.Asafieldofacademicstudy,manyAIresearchersreachtounderstandintelligencebybecomingabletoproduceeffectsofintelligence:intelligentbehavior.OneelementinAI’smethodologyisthatprogressissoughtbybuildingsystemsthatperform:synthesisbeforeanalysis[78].“Systemsaregoodscience”,asHendlersaid[34].OrmoredrasticallybyWachsmuth[78]:“itisnottheaimofAItobuildintelligentmachineshavingunderstoodnaturalintelligence,buttounderstandnaturalintelligencebybuildingintelligentmachines”.EvenmorestrikinglyAaronSlomanputsitthisway(bycitinghiscolleagueRusselBeale):AIcanbedefinedastheattempttogetrealma-chinestobehaveliketheonesinthemovies”.Inaddition,hepointsoutthatAIhastwomainstrands,ascientificstrandandanengineeringstrand,whichoverlapconsidera-blyintheirconcepts,methods,andtools,thoughtheirobjec-tivesareverydifferent.ThisviewissupportedbyWahlster[79]whoclarifiesthatAIhastwodifferenttypesofgoals,onemotivatedbycogni-tivescience,theotherbytheengineeringsciences(cf.Figure1).2Figure1AIandRelatedResearchAreas(adaptedfrom[79])Afurthersub-division(adaptedfromRichter[59]andAbecker[4])ofAIintosub-fields,methods,andtechniquesisshowninFigure2.Figure2AIFields,Methods,andTechniquesForSEthescientificstrandorientatingtowardscognitivescienceandhumanitiesingeneralcouldbeahelpfulguid-anceforinterdisciplinaryresearch.Ofcourse,thereisastrongoverlapbetweenSEandtheengineeringstrandofAI.AnimportantpartofthelatterareKBS.Richter[59]definesthreedifferentlevelsasessentialfordescribingKBS:thecognitivelayer(human-oriented,ra-tional,informal),therepresentationlayer(formal,logical),andtheimplementationlayer(machine-oriented,datastruc-turesandprograms).TheselevelsareshowninFigure3.Betweentheknowledgeutteranceanditsmachineutiliza-tionseveraltransformationshavetobeperformed(thickarrows).Theypointtothedirectionofincreasedstructuringwithinthelayersandproceedfromthecognitiveformtoamoreformalandmoreefficientlyprocessedform.TheletterAisareminderforAcquisition(whichishuman-oriented)whileCisashorthandforCompilation(machine-oriented).Eachsyntacticresultintherangeofatransformationbe-tweenlayershastobeassociatedwiththemeaninginthedomainofthetransformation.Themostinterestinganddifficulta