FORMALIZED QUALITATIVE MODELING OF ONLINE TRUST: INTRODUCTION OF THE METHOD AND A DETAILED EXAMPLE

201 2, XIX, 2016 In o ma ion Managemen DOI: 10.15240/ ul/001/2016-2-014 In oduc ion The ask o he p esen pape is o ou line a o malized quali a i e model (FQM) o online us (OT). OT is i al o es ablishing and main aining comme cial ela ionships on he in e ne (G abne -K aeu e , 2002; Van De Heijden e al., 2003; Kim e al., 2008; 2012; Beldad e al., 2010; Delina & D áb, 2010). T us is especially impo an when he e is ela i e lack o ce ain y, o mal ules, cus oms and ag eemen s (B adach e al., 1989; Ge en & S aub, 2004), as is o en he case wi h e-comme ce in e ac ion and online in e ac ion in gene al (Ge en, 2000; McKnigh e al., 2002; Ge en & S aub, 2004). OT is a mul idimensional p oblem (McKnigh & Che any, 2001; McKnigh e al., 2002; Shanka e al., 2002; Chau e al., 2007; Beldad e al., 2010). FQMs a e well sui ed o ackling mul idimensional asks. Also, FQMs a e use ul whene e ague and/o quali a i e in o ma ion need o be included in o a model. In he OT con ex i migh be necessa y o conside o ins ance he in l uence o pe cei ed social p esence embedded in a web si e (Ge en & S aub, 2004; Cy e al., 2007; Hassanein & Head, 2007) o posi i e wo d o mou h (G abne - K aeu e , 2002; Co bi e al., 2003; U z e al., 2012). I migh be di i cul o p ohibi i ely cos ly o measu e hese in l uences p ecisely o quickly enough (e.g. when online companies need o mee deadlines o s a egic decisions). I is app op ia e o use an FQM hen. FQMs cap u e ela ionships among a iables in he o m o deg aded (simpli i ed) equa ions and s a is ical ela ions and/o in he o m o common-sense heu is ics (e.g. i X goes up, Y goes down wi h inc easing apidi y). Quali a i e me hodology (Kuipe s, 1989; Dohnal, 1991; T a e-Massuyes e al., 2004) has been used in some o m o model o example in es men decisions and economic p oblems (Bena och & Dha , 1995; Hinkkanen e al., 2003; Cu ic e al., 2008; Konečný e al., 2010; Kocmano á e al., 2011) and a a ie y o enginee ing p oblems. See Bou seau e al. (1995), De Jong (2004) and P ice e al. (2006) o an o e iew. Quali a i e models cap u e he undamen al ea u es o a sys em unde s udy, while elimina ing quan i a i e de ail (Kuipe s, 1989). Quali a i e modeling can be seen as one o he “unce ain y calculi”, such as uzzy se s (Zadeh, 1965; Dubois & P ade, 1991; Hub & Za loukal, 2009), ough se s (Pawlak, 1982) and o de o magni ude easoning (Raiman, 1991). Such calculi can be help ul when dealing wi h online us (Song e al., 2005; Chak abo y & Chak abo y, 2007; Li e al., 2009; Li e al., 2012), especially unde in o ma ion sho age, measu emen di i cul ies, ime p essu e o make decisions and/o unce ain y, o when se e al no el, subjec i e and/o di i cul o measu e (e.g. quali a i e only) a iables a e being conside ed. We p opose a gene al me hodological amewo k ha enables inco po a ion o many e en e y ague and e y di e se in l uences on OT in he con ex o e-comme ce. Al hough he speci i c e sion o he quali a i e algo i hm o iginally p oposed by he second au ho has been used se e al imes in he pas , mos ecen ly in Vícha & Dohnal (2008a, 2008b), Konečný e al. (2010), Kocmano á e al. (2011), Režňáko á e al. (2012), his is ac ually he i s pape whe e one o he p incipal aspec s o he algo i hm, namely selec ion o a consis en se o scena ios, is ea ed explici ly and in de ail (see sec ion 1.2). 1. Me hod 1.1 Quali a i e Models The e a e only h ee quali a i e alues: posi i e, ze o and nega i e. The symbols used FORMALIZED QUALITATIVE MODELING OF ONLINE TRUST: INTRODUCTION OF THE METHOD AND A DETAILED EXAMPLE Š ěpán Veselý, Mi ko Dohnal EM_2_2016.indd 201EM_2_2016.indd 201 3.6.2016 11:48:123.6.2016 11:48:12 202 2016, XIX, 2 In o mační managemen a e +, 0, -, espec i ely. A quali a i e scena io o a quali a i e model is speci i ed i all i s n quali a i e a iables X  (X1, X2, …, Xn) a e desc ibed by he quali a i e iple s (X, DX, DDX), whe e DX and DDX a e he i s and second quali a i e de i a i es wi h espec o ime (o an independen a iable in gene al). Le us suppose ha he iple (++0)  (T, DT, DDT) ep esen s isk a e sion R( ) as a unc ion o ime: i means ha isk a e sion is posi i e (R = +), i is inc easing in his example (DR = +) and he inc ease is linea (DDR = 0) as he second de i a i e is ze o. In o mally speaking, he i s quali a- i e de i a i e ells us whe he a a iable is inc easing, dec easing, o s able as a unc ion o ano he a iable – he i s quali a i e de i a i e is posi i e, nega i e, and ze o in hese cases, espec i ely. Again in o mally speaking, he second quali a i e de i a i e ells us whe he such a change (i.e. inc ease o dec ease) in a a iable is accele a ing, s able o decele a ing – he second quali a i e de i a i e is posi i e, ze o, and nega i e in hese cases, espec i ely. Looking a Fig. 1 on he nex page, in pic u es 21, 22 and 23 a e ins ances o unc ions wi h posi i e i s quali a i e de i a i e, while in pic u es 24, 25 and 26 we can see ins ances o unc ions wi h nega i e i s quali a i e de i a i e. A unc ion wi h he i s quali a i e de i a i e equal o ze o would be ep esen ed by a ho izon al s aigh line (pa allel wi h he X axis). Pic u es 22 and 25 in Fig. 1 gi e examples o unc ions wi h he second quali a i e de i a i e equal o ze o. Pic u es 21 and 26 ep esen unc ions wi h posi i e second quali a i e de i a i e ( he inc ease o dec ease in Y is becoming mo e and mo e p onounced as a unc ion o X), while pic u es 23 and 24 ep esen unc ions wi h nega i e second quali a i e de i a i e (we can see ha he inc ease o dec ease in Y wi h espec o he inc ease in X g adually l a ens ou ). A ypical example o a quali a i e knowledge i em can be o malized by a ce ain simple ela ion be ween wo a iables X and Y. Fo example: I he p ice X o a p oduc is dec easing hen he demand Y is inc easing. (1) A o mal in e p e a ion o he quali a i e knowledge i em (1) is DY/DX = -, whe e DY/DX is he i s quali a i e de i a i e o Y wi h espec o X (in his example he de i a i e is nega i e). Typical examples o quali a i e ela ions a e gi en in Fig. 1. The iden i i ca ion numbe s gi en in Fig. 1 a e shape codes o he espec i e quali a i e shapes, i.e. o ins ance 21 is a code numbe o unc ion cha ac e ized by posi i e alue o Y and posi i e i s and second quali a i e de i a i es o Y wi h espec o X ( iple +++). The iden i i ca ion numbe s a e employed in sec ion 2.1. I he second de i a i e is no known hen he e a e wo a ian s o quali a i e p opo ionali y: M_+ I X is inc easing hen Y is inc easing. I X is dec easing hen Y is dec easing. (2) M_- I X is inc easing hen Y is dec easing. I X is dec easing hen Y is inc easing. (3) Fo mo e de ails see e.g. Kuipe s (1989), Pa sons and Dohnal (1995), T a e-Massuyes e al. (2004). A key concep in he app oach o quali a i e modeling p esen ed he e is “quali a i e scena io” o simply “scena io”. In plain wo ds, a scena io ep esen s a concise quali a i e desc ip ion e.g. o an objec o a sys em. An example o a scena io could be: I p ice o p oduc X inc eases, we will sell less X. Ano he example could be: I p ice o X is s able and i X’s epu a ion o quali y inc eases, we will sell mo e X. These wo examples desc ibe a business de elopmen using wo (p ice and sales) and h ee (p ice, sales and epu a ion o quali y) a iables, espec i ely. Technically, a scena io consis s o a se ies o a iables. Each a iable in he scena io is quali a i ely desc ibed by i s s a e (posi i e quan i y, nega i e quan i y, ze o quan i y) and by he i s and second quali a i e de i a i es (usually wi h espec o ime). As we al eady s a ed a he beginning o his sec ion, hese h ee quali ies each a iable can ake can ha e only h ee possible alues (+, - and 0). The h ee quali ies each a iable can ake can he e o e be desc ibed by a sequence o h ee signs, i.e. a iple , as was also al eady men ioned abo e. To illus a e how a scena io is desc ibed using iple s, le us e u n o he example o a quali a i e scena io men ioned p e iously: “I p ice o X is s able and i X’s epu a ion EM_2_2016.indd 202EM_2_2016.indd 202 3.6.2016 11:48:123.6.2016 11:48:12 203 2, XIX, 2016 In o ma ion Managemen o quali y inc eases [le us assume ha i inc eases linea ly, so he second quali a i e de i a i e is ze o], we will sell mo e X [le us assume ha he inc ease in sales is slowing down because o g adual demand sa u a ion, so he second quali a i e de i a i e is nega i e].” This scena io can be desc ibed by he ollowing se ies o iple s: +00 (p ice), ++0 ( epu a ion), ++- (sales). Each scena io can be con enien ly displayed in one ow o a able whe e each cell con ains quali a i e desc ip ion o one a iable/ one iple (see Tab. 2 in sec ion 2.3). As we will show nex , some scena ios ep esen possible solu ions o a quali a i e model because hey a e consis en wi h a se o cons ain s based on a ailable knowledge abou a p oblem (e.g. abou online us and i s de e minan s), while o he scena ios a e disca ded because hey a e no consis en wi h he se o cons ain s (see sec ion 1.2). Those disca ded scena ios ypically include ce ain ela ions be ween a iables ha a e deemed unlikely o impossible based on he a ailable knowledge ( o example a nega i e ela ion be ween he le el o us owa ds an online si e and his si e’s pe cei ed quali y and ease o use). 1.2 Quali a i e Vec o Op imiza ion Le us suppose ha he e a e wo independen a iables X1, X2 and wo objec i e unc ions Q1, Q2. The e is a ec o F o cons ain s ep esen ed Fig. 1: Examples o pai -wise quali a i e equa ionless ela ions Sou ce: own EM_2_2016.indd 203EM_2_2016.indd 203 3.6.2016 11:48:123.6.2016 11:48:12 204 2016, XIX, 2 In o mační managemen by a se o equa ionless ela ions among (some o all) o he a iables X1, X2, Q1, Q2: F(X1, X2, Q1, Q2) = 0 (4) Such a se o cons ain s (4) cons i u es he quali a i e model o he p oblem a hand. A gi en model can ha e any numbe o independen a iables and objec i e unc ions. A conc e e example o a quali a i e model (i.e. a se o cons ain s) desc ibing he p oblem o online us is gi en in (6) in sec ion 2.2. Ano he e y simple example o a cons ain is gi en in (5) in his sec ion. The cons ain (5) consis s o a single ela ion (namely he nega i e quali a i e p opo ionali y) be ween wo a iables, X and Y. This e y simple model (5) will be used o illus a e how he cons ain s a e employed o di e en ia e solu ions o a p oblem ha a e consis en and solu ions ha a e inconsis en wi h a model o a p oblem. In simple wo ds we can say ha solu ions consis en wi h he model can happen, while inconsis en solu ions can no ake place in eali y (as a as he knowledge used o cons uc he model is alid, o cou se). A gi en model (i.e. he se o cons ain s) can be based on empi ical obse a ion, on p e iously published esul s o on expe opinions. The p esen quali a i e algo i hm is based on sys ema ic con on a ion o all possible iple s o each a iable and he model i sel . Scena ios (solu ions) consis en wi h he model (i.e. wi h he se o cons ain s) a e selec ed, while inconsis en scena ios a e ejec ed. This ype o solu ion is called b u al o ce in a i i cial in elligence. We ha e now desc ibed, in gene al e ms, how a quali a i e model is sol ed. To ecapi ula e: he possible solu ions (i.e. scena ios) a e selec ed i hey a e consis en wi h a se o cons ain s o med by a se o quali a i e ela ions be ween a iables (such as (5) and (6) below). In he ollowing pa ag aphs, we desc ibe he p ocess o sol ing a quali a i e model in mo e de ail. Reade s in e es ed mainly in he applica ion o quali a i e modeling o online us can now skip di ec ly o sec ion 2. As we al eady unde s and on he concep ual le el, scena ios selec ion can be seen as a consis ency p oblem. A he beginning o he p ocess o model de elopmen we usually ha e jus a se o a iables. E en a his ini ial s age i is possible o calcula e solu ions o he quali a i e model. Bu i would no ha e any p ac ical alue, because we would ob ain all “imaginable” scena ios, i.e. all combina ions o quali a i e alues o each iple . Recall ha in a scena io each a iable is ep esen ed by a quali a i e iple (X, DX, DDX) – see sec ion 1.1. In he p esen model o online us we ha e 13 a iables (see sec ion 2.1), hence 13 iple s in each scena io. Each alue (X, DX and DDX) in he iple can be ei he +, -, o 0. So, a his ini ial s age, he e a e 33 = 27 possible combina ions o each iple . Since we ha e 13 such iple s (one o each a iable) in each scena io, he e a e 2713 possible combina ions o +, -, and 0, whe e each combina ion ep esen s one “imaginable” scena io. This as numbe o 2713 scena ios needs o be educed o ob ain a p ac ical solu ion. This is done by including new knowledge in he model. Be o e we show how he numbe o scena ios is educed o he online us model (see Tab. 3 in sec ion 2.3), we will ou line he basic p inciple o how inconsis en scena ios a e disca ded om he model wi h he inclusion o knowledge i ems. Conside he ollowing simple example. The in e ela ion M_- X Y (5) s a es he e is a nega i e ela ionship be ween a iables X and Y. All scena ios ha iola e his ela ionship mus be disca ded om he model when his knowledge i em is en e ed. I we had a model wi h jus wo a iables (X and Y), he model solu ion be o e and a e en e ing in e ela ion (5) would look as shown in Tab. 1. The solu ion in Tab. 1 is in ui i ely comp ehensible: we mus simply exclude all scena ios ha ha e o he alues o i s de i a i es (DX) han ei he “+” o X and “-“ o Y o “-“ o X and “+” o Y. To see his, le us say ha X is inc easing as a unc ion o some a iable Z (e.g. ime). This means ha he i s quali a i e de i a i e o X wi h espec o Z is posi i e. I du ing he same change in Z (e.g. ime) Y also inc eased, i.e. DY = + (wi h espec o Z), his would mean ha X and Y mo ed in he same di ec ion (bo h inc eased as a unc ion o Z). This is ep esen ed by scena io 1 in Tab. 1. Howe e , such a de elopmen is no possible gi en cons ain (5) which equi es a nega i e p opo ionali y be ween X and Y, i.e. when X goes up, Y mus go down and ice EM_2_2016.indd 204EM_2_2016.indd 204 3.6.2016 11:48:123.6.2016 11:48:12 205 2, XIX, 2016 In o ma ion Managemen e sa. Consequen ly, only scena ios 5 and 7 in Tab. 1 a e consis en wi h cons ain (5). In con as , scena ios 1 and 2 (in Tab. 1) would be consis en wi h a cons ain s ipula ing a posi i e p opo ionali y be ween he a iables. Scena ios 4, 6, 8 and 9 indica e he e is no ela ion be ween he a iables (when one a iable changes in ei he di ec ion, he e is no change in he emaining a iable). Scena io 3 desc ibes a si ua ion whe e he e a e no changes in ei he a iable, hus i canno be shown ha he e ac ually is a posi i e ela ion be ween X and Y as equi ed by (5). Scena io 3 is he e o e also disca ded om he se o solu ions consis en wi h cons ain (5). 2. Resul s and Discussion 2.1 Model Va iables OT can be cha ac e ized by he ollowing se o 13 a iables, 11 independen a iables and wo objec i e unc ions: Objec i e unc ions: TRU Le el o us owa ds an online si e/ company RIS Pe cei ed isk o in e ac ion wi h he si e/company (Ho man e al., 1999; Yoon, 2002; Lacohee e al., 2006; Kim e al., 2008) Independen a iables: FAM Familia i y wi h he si e/company (Ge en, 2000; Ge en & S aub, 2004) MAR Le el o he online company’s ma ke o ien a ion, e.g. se ices cus omiza ion (S ini asan e al., 2002; Co bi e al., 2003; Koehn, 2003) SIT Pe cei ed websi e quali y and ease o use (McKnigh & Che any, 2001; G abne -K aeu e , 2002; Liao e al., 2006; Chau e al., 2007; Kim e al., 2008, 2012) SOC Social p esence embedded in he web si e (Ge en & S aub, 2004; Cy e al., 2007; Hassanein & Head, 2007) BEN Pe cei ed bene olence o he online company (McKnigh e al., 2002; Ge en & S aub, 2004) REP Si e’s/company’s posi i e epu a ion (McKnigh & Che any, 2001; Casalo e al., 2008; Kim e al., 2008) WOM Le el o posi i e o wo d o mou h (G abne -K aeu e , 2002; Co bi e al., 2003; Lacohee e al., 2006) PAR Si e’s/company’s pa ne ship wi h well known pa ne s (Co bi e al., 2003; Zhang, 2004; Kim e al., 2008; Hong & Cho, 2011) EXP Use ’s web expe ience (McKnigh e al., 2002; Co bi e al., 2003; Lacohee e al., 2006; Me zge , 2006; Li e al., 2009) Model be o e en e ing in e ela ion (5) Model a e en e ing in e ela ion (5) Va iables Va iables Scena io X Y Scena io X Y 1 ++* ++* 2 +-* +-* 3 +0* +0* 4 ++* +0* 5 +-* ++* 5 +-* ++* 6 +0* +-* 7 ++* +-* 7 ++* +-* 8 +-* +0* 9 +0* ++* No e: All a iables a e assumed o be posi i e. Fo simplici y, second de i a i es a e no conside ed in his example, hus he second de i a i e (DDX) is deno ed by * in all iple s (* can mean ei he +, - o 0). Sou ce: own Tab. 1: Scena ios be o e and a e en e ing an in e ela ion EM_2_2016.indd 205EM_2_2016.indd 205 3.6.2016 11:48:133.6.2016 11:48:13 206 2016, XIX, 2 In o mační managemen PRE Numbe o he use ’s p e ious posi i e expe iences wi h online in e ac ion (Ge en & S aub, 2004; Fla ian e al., 2006; Casalo e al., 2008) TEC Pe cei ed echnological us wo hi- ness o he si e/company (G abne - K aeu e , 2002; Co bi e al., 2003; Lacohee e al., 2006) 2.2 Model Ins uc ions The ollowing se (6) o pai -wise quali a i e ela ions is used o o malize ela ions among he se o a iables om sec ion 2.1. The se o in e ela ions is inspi ed by many dialogues wi h a eam o expe s and a g oup o in e ne use s and by s udies quo ed abo e in sec ion 2.1. The expe s we e MBA s uden s a B no Uni e si y o Technology wi h subs an ial expe ience in online ade and/o ma ke ing. The in e ne use s we e selec ed p e-g adua e s uden s a B no Uni e si y o Technology. The quali a i e model is ep esen ed by he ollowing se o ela ions (see Fig. 1): No. Shape X Y (see Fig. 1) 1 23 WOM FAM 2 M_+ (see (2)) SIT MAR 3 M_+ REP MAR 4 21 WOM MAR 5 21 WOM SIT 6 21 TEC SIT 7 21 BEN SOC 8 M_+ REP BEN 9 M_+ WOM BEN 10 23 PAR BEN 11 23 PRE BEN 12 M_+ TEC BEN 13 23 WOM REP 14 M_+ TEC REP 15 M_+ TEC EXP 16 26 RIS TRU 17 23 FAM TRU 18 23 MAR TRU 19 23 SIT TRU 20 23 SOC TRU 21 23 BEN TRU 22 23 REP TRU 23 23 PAR TRU 24 M_+ PRE TRU 25 24 FAM RIS 26 24 SIT RIS 27 M_- BEN RIS 28 24 REP RIS 29 24 PAR RIS 30 24 EXP RIS 31 M_- PRE RIS 32 24 TEC RIS (6) 2.3 Model Resul s – Scena ios The se o 23 scena ios – see Tab. 2 – is gene a ed using so wa e employed in Vícha and Dohnal (2008a; 2008b). The so wa e was p og ammed by a g oup a ound he second au ho . As a as we know, he e a e cu en ly no widesp ead comme cial so wa es o analyzing quali a i e p oblems and di e en esea ch g oups o en use hei own so wa es. Howe e , analy ical ools o quali a i e compu a ions can be p og ammed o example in MATLAB. Di e en quali a i e p oblems ela ed o online us can be easily sol ed using he se o scena ios in Tab. 2 which ep esen a comple e desc ip ion o all possible beha io s wi hin he modeled sys em (cons ained by ela ions gi en in (6)). All a iables in Tab. 2 a e posi i e because o hei e y na u e. The e o e he i s alues in all iple s (in Tab. 2) a e always equal o +. Fo example scena io 12 is a s eady s a e si ua ion: all i s and second de i a i es a e ze os. The e o e no hing is happening, he e a e no changes in ime. Scena ios 1–10 indica e e.g. ha le el o us owa ds he company (TRU) inc eases as a unc ion o ime and pe cei ed isk o in e ac ion wi h he si e/company (RIS) dec eases as a unc ion o ime, while he es o he a iables go up. Scena ios 14–23 gi e he opposi e. Tha means all independen a iables in he p esen model a e posi i ely linked o TRU and nega i ely linked o RIS ( his is no su p ising, e.g. in Kim and Pa k (2013) six ou o se en independen a iables we e posi i ely ela ed o online us ). The e a e, howe e , some di e ences in he p ecise cha ac e o he inc ease/decline o he a iables wi h espec o he second de i a i es. Fo example in scena io 1 all a iables inc ease (o in he case o RIS dec ease) mo e and mo e quickly (all iple s a e equal o +++ and +--, espec i ely), whe eas in scena io 10 all a iables inc ease (o in he case o RIS dec ease) wi h a dec easing apidi y (all iple s a e equal o ++- and +-+, espec i ely). EM_2_2016.indd 206EM_2_2016.indd 206 3.6.2016 11:48:133.6.2016 11:48:13 207 2, XIX, 2016 In o ma ion Managemen The na u e o one o he objec i e unc ions (TRU) equi es maximiza ion, whe eas he o he objec i e unc ion (RIS) equi es minimiza ion, hence, he e a e 10 ela i ely a o able scena ios, i.e. scena ios 1–10, and 10 ela i ely undesi able ones, scena ios 14–23. Le us o example suppose ha an online company is unce ain abou he e ec an inc ease in he a iable “social p esence embedded in he web si e” (SOC) will ha e on “pe cei ed isk o in e ac ion wi h he si e/company” (RIS). Ha ing a look a scena ios 1–10 (and columns RIS and SOC) in Tab. 2 would ell he company manage s ha he e ec ough o be posi i e. Also, i seems ha he e ec would be almos always he same (wi h he excep ion o scena io 1) i espec i e o he speed wi h which he changes in SOC a e implemen ed. We need o keep in mind, hough, ha FQMs a e in o ma ion non-in ensi e. This implies ha hey usually cap u e only he mos obus aspec s o he desc ibed sys ems. This is he case especially when he solu ions a e na owed down o jus a ew scena ios (i.e. when he amoun o a ailable quali a i e knowledge is la ge, see Tab. 3 and ex below). Mo e in o ma ion in ensi e me hods (e.g. uzzy ma hema ical me hods o s a is ics) can be subsequen ly used o gain addi ional insigh s in o he s udied sys ems/ p oblems. Howe e , his is no always an op ion in high-speed en i onmen s, o example because collec ion o quan i a i e da a akes ime. Scena io Va iables (see sec ion 2.1) TRU RIS FAM MAR SIT SOC BEN REP WOM PAR EXP PRE TEC 1 +++ +-- +++ +++ +++ +++ +++ +++ +++ +++ +++ +++ +++ 2 ++- +-+ ++- ++- ++- +++ ++- ++- ++- +++ ++- ++- ++- 3 ++- +-+ ++- ++- ++- +++ ++- ++- ++- ++0 ++- ++- ++- 4 ++- +-+ ++- ++- ++- +++ ++- ++- ++- ++- ++- ++- ++- 5 ++- +-+ ++- ++- ++- ++0 ++- ++- ++- +++ ++- ++- ++- 6 ++- +-+ ++- ++- ++- ++0 ++- ++- ++- ++0 ++- ++- ++- 7 ++- +-+ ++- ++- ++- ++0 ++- ++- ++- ++- ++- ++- ++- 8 ++- +-+ ++- ++- ++- ++- ++- ++- ++- +++ ++- ++- ++- 9 ++- +-+ ++- ++- ++- ++- ++- ++- ++- ++0 ++- ++- ++- 10 ++- +-+ ++- ++- ++- ++- ++- ++- ++- ++- ++- ++- ++- 11 +0+ +0- +0+ +0+ +0+ +0+ +0+ +0+ +0+ +0+ +0+ +0+ +0+ 12 +00 +00 +00 +00 +00 +00 +00 +00 +00 +00 +00 +00 +00 13 +0- +0+ +0- +0- +0- +0- +0- +0- +0- +0- +0- +0- +0- 14 +-+ ++- +-+ +-+ +-+ +-+ +-+ +-+ +-+ +-+ +-+ +-+ +-+ 15 +-- +++ +-- +-- +-- +-+ +-- +-- +-- +-+ +-- +-- +-- 16 +-- +++ +-- +-- +-- +-+ +-- +-- +-- +-0 +-- +-- +-- 17 +-- +++ +-- +-- +-- +-+ +-- +-- +-- +-- +-- +-- +-- 18 +-- +++ +-- +-- +-- +-0 +-- +-- +-- +-+ +-- +-- +-- 19 +-- +++ +-- +-- +-- +-0 +-- +-- +-- +-0 +-- +-- +-- 20 +-- +++ +-- +-- +-- +-0 +-- +-- +-- +-- +-- +-- +-- 21 +-- +++ +-- +-- +-- +-- +-- +-- +-- +-+ +-- +-- +-- 22 +-- +++ +-- +-- +-- +-- +-- +-- +-- +-0 +-- +-- +-- 23 +-- +++ +-- +-- +-- +-- +-- +-- +-- +-- +-- +-- +-- Sou ce: Au ho s Tab. 2: Online us scena ios EM_2_2016.indd 207EM_2_2016.indd 207 3.6.2016 11:48:133.6.2016 11:48:13 208 2016, XIX, 2 In o mační managemen We can say ha he p esen model is well speci i ed by in e ela ions gi en in (6), which causes i o be qui e es ic i e, i.e. he e is li le a iabili y in he scena ios ob ained. I less inpu knowledge (such as (6)) was a ailable, he model would be less well speci i ed and consequen ly less es ic i e, and mo e a iabili y would be encoun e ed in he scena ios ob ained (e.g. ce ain a iables would be linked o ce ain o he a iables posi i ely in a gi en subse o scena ios and nega i ely in ano he subse ). Relying on a es ic i e model has a po en ial disad an age in ha ce ain sub le ela ions o di e ences may be dis ega ded in he model in o de o ob ain he se o he mos obus o mos ypical scena ios (e.g. in he p esen model we igno e he possibili y ha he ela ion be ween SIT and TRU migh be pa ially cul u e-sensi i e, see Pa k e al., 2012). On he o he hand, i is also possible o de elop an ad hoc quali a i e model “ ypical” o an unusual si ua ion/sys em. Table 3 displays he numbe o online us scena ios ob ained using he basic p inciple o consis ency ou lined in sec ion 1.2. The columns “In e ela ions en e ed” gi e he ange o in e ela ions en e ed (i.e. 1–3 means ha in e ela ions 1, 2 and 3 om (6) ha e been en e ed in o he model). Using he p inciple o consis ency, he as numbe o “imaginable” scena ios (2713) has been apidly educed by en e ing a ailable pieces o knowledge. As is appa en om Tab. 3 he solu ions o quali a i e models become in e p e able a e including abou 25 pai -wise in e ela ions be ween a iables. This s a emen is ue o models wi h abou he same numbe o a iables we ha e in ou model o online us . Mo e (less) knowledge i ems need o be en e ed i he model ea u es mo e (less) a iables. I can also be obse ed ha en e ing ce ain in e ela ions (e.g. in e ela ions numbe 5, 13, 14, 18, 26 – see Tab. 3) does no educe he numbe o scena ios ob ained. The eason is ha he espec i e se o scena ios be o e he pa icula in e ela ion was en e ed had al eady been consis en wi h ha in e ela ion. FQM de elopmen is a mul i-s ep p ocess, and he p esen model can be modi i ed/ upg aded o add ess speci i c p ac ical and In e ela ions en e ed Numbe o scena ios In e ela ions en e ed Numbe o scena ios None 2713 1-16 819 All a iables posi i e* 913 1-17 187 1 407953774917 1-18 187 1-2 45328197213 1-19 187 1-3 5036466357 1-20 187 1-4 731794257 1-21 187 1-5 731794257 1-22 187 1-6 176969853 1-23 187 1-7 28402569 1-24 59 1-8 2421009 1-25 23 1-9 1594323 1-26 23 1-10 255879 1-27 23 1-11 54675 1-28 23 1-12 45927 1-29 23 1-13 45927 1-30 23 1-14 45927 1-31 23 1-15 5103 1-32 23 No e: * We assume ha all a iables a e posi i e la e on (in his able) as well. Sou ce: own Tab. 3: Numbe o scena ios dependen on he pai -wise ela ions en e ed in he model EM_2_2016.indd 208EM_2_2016.indd 208 3.6.2016 11:48:133.6.2016 11:48:13 209 2, XIX, 2016 In o ma ion Managemen heo e ical needs. Also, i is possible o calcula e all possible ansi ions be ween he quali a i e scena ios (Režňáko á e al., 2012). The possibili y o upda e quali a i e models quickly and easily based on an in l ow o new quali a i e in o ma ion can be an ad an age in u bulen en i onmen s such as elec onic e ail ma ke s o he compu e and mobile phones indus ies. The ob ious limi a ion o FQM is ha quali a i e easoning can answe quali a i e que ies only, no quan i a i e ones, and hus can se e jus as a complemen o he exis ing quan i a i e me hods (and/o as a igo ous o maliza ion o quali a i e me hods, such as case s udies). Howe e , inclusion o ma hema ical equa ions in o quali a i e models is possible (Vícha & Dohnal, 2008a; 2008b). FQMs can en ich he me hodological ins umen a ium used in he s udy o OT (e.g. G abne -K au e & Kaluscha (2003) ad oca e he use o a b oad spec um o esea ch ools). The main ad an ages o FQMs a e:  No nume ical alues o cons an s and pa ame e s a e needed (i.e. in o ma ion non-in ensi e).  The se o possible solu ions (scena ios) is comple e, i.e. he e canno be any o he quali a i e scena ios ha a e no gene a ed by he quali a i e model.  FQMs a e l exible, di e se a iables can be included.  They a e inexpensi e and eadily unde s andable by p ac i ione s. Conclusion Quali a i e app oach has much o o e when highly complex and/o pa ially ague p oblems such as OT a e examined. In he p esen s udy o OT de e minan s a o mal ool o dealing wi h da a o non-nume ical na u e was employed o gene a e a FQM consis ing o 23 possible scena ios. The model ob ained is de i ni ely no he only possible al e na i e. Many modi i ca ions, upg ades and ex ensions a e possible. The pape p esen s jus me hodology and a simple model as a demons a ion. Also, his is he i s pape whe e one o he p incipal aspec s o he p esen quali a i e algo i hm, namely selec ion o a consis en se o scena ios, is ea ed explici ly and in de ail. FQMs can complemen es ablished ools o OT analysis wi h e y li le addi ional cos . This migh be p o i able especially unde in o ma ion sho age, measu emen di i cul ies, ime p essu e o make decisions and/o unce ain y, o when se e al no el, subjec i e and/o di i cul o measu e (e.g. quali a i e only) a iables a e conside ed when dealing wi h online us in he e-comme ce con ex . Re e ences Beldad, A., De Jong, M., & S eehoude , M. (2010). How shall I us he aceless and he in angible? A li e a u e e iew on he an eceden s o online us . Compu e s in Human Beha io , 26(5), 857-869. doi:10.1016/j. chb.2010.03.013. Bena och, M., & Dha , V. (1995). Con olling he complexi y o in es men decisions using quali a i e easoning echniques. Decision Suppo Sys ems, 15(2), 115-131. doi:10.1016/0167-9236(94)00031-M. Bou seau, P., Bousson, K., Dague, P., Do moy, J.L., E a d, J.M., Gue in, F., Ley al, L., Lhomme, O., Lucas, B., Missie , A., Mon main, J., Pie a, N., Rako o-Ra alon - salama, N., S eye , J.P., Tomasena, M., T a e- Massuyes, L., Vesco i, M., Xan hakis, S., & Yannou, B. (1995). Quali a i e easoning: A su ey o echniques and applica ions. AI Communica ions, 8(3-4), 119-192. doi:10.3233/ AIC-1995-83-401. B adach, J.L., & Eccles, R.G. (1989). P ice, au ho i y, and us : F om ideal ypes o plu al o ms. Annual Re iew o Sociology, 15, 97-118. doi:10.1146/annu e .so.15.080189.000525. Casalo, L., Fla ian, C., & Guinaliu, M. (2008). The ole o pe cei ed usabili y, epu a ion, sa is ac ion and consume amilia i y on he websi e loyal y o ma ion p ocess. Compu e s in Human Beha io , 24(2), 325-345. doi:10.1016/j.chb.2007.01.017. Chak abo y, C., & Chak abo y, D. (2007). Fuzzy ule base o consume us wo hiness in In e ne ma ke ing: An in e ac i e uzzy ule classi i ca ion app oach. In elligen Da a Analysis, 11(4), 339-353. doi:10.1016/j. dss.2003.12.005. Chau, P.Y.K., Hu, P.J.-H., Lee, B.L.P., & Au, A.K.K. (2007). Examining cus ome s’ us in online endo s and hei d opou decisions: An empi ical s udy. Elec onic Comme ce Resea ch and Applica ions, 6(2), 171-182. doi:10.1016/j.ele ap.2006.11.008. Co bi , B.J., Thanasanki , T., & Yi, H. (2003). T us and e-comme ce: A s udy o consume pe cep ions. Elec onic Comme ce Resea ch EM_2_2016.indd 209EM_2_2016.indd 209 3.6.2016 11:48:133.6.2016 11:48:13