B usi ness Pr oc ess Ver i fi c ati on wi th T empor al - - PowerPoint PPT Presentation

B usi ness Pr

• c ess Ver

i fi c ati

• n wi

th T empor al Answer S et Pr

• gr

ammi ng

L a ura Gio rda no , Alb e rto Ma rte lli, Ma tte o Spio tta , Da nie le T he se ide r Dupré Unive rsità de l Pie mo nte Orie nta le Unive rsità di T

• rino

I ta ly

Go a l

Cro ss-fe rtilize Busine ss Pro c e ss Mo de ling with c o ntrib utio ns fro m Re asoning about ac tions and

c hange in AI a nd Answe r Se t Pr

• gr

amming

• De c la ra tive o r pro c e dura l pro c e ss mo de l
• Mo de ling dire c t e ffe c ts o f a c tivitie s a s we ll a s

side e ffe c ts with c ausal r

ule s (b a c kg ro und

kno wle dg e )

• F

le xib le mo de ling o f o b lig a tio ns

• Mo de ling da ta
• Ve rific a tio n o f c o mplia nc e with no rms a nd

b usine ss rule s

Our c o ntrib utio n

• De c la ra tive / Pro c e dura l de sc riptio n o f Pro c e sse s
• Ac tio n e ffe c ts
• No rms

e xpre sse d in a T

e mpor al e xte nsio n o f Answe r Se t Pr

• gr

amming

(a no nmo no to nic kno wle dg e re pre se nta tio n & re a so ning fra me wo rk) Ve rific a tio n o f te mpo ra l lo g ic pro pe rtie s pe rfo rme d a s Bounde d Mode l Che c king in T e mpo ra l ASP

T he I CT 4L AW pro je c t

“I CT Co nve rg ing o n L a w: Ne xt Ge ne ra tio n Se rvic e s fo r Citize ns, E nte rprise s, Pub lic Administra tio n a nd Po lic yma ke rs” 2009-2013, funde d b y Re g io ne Pie mo nte Se ve ra l pa rtne rs, inc luding , fo r pro c e dure c o mplia nc e with no rms:

• I

T T I G a nd L OA (CNR)

• Aug e o s, SSB Pro g e tti (priva te c o mpa nie s)
• Our unive rsitie s

Ac tio n the o rie s

F lue nts: the ir truth va lue de sc rib e s the sta te o f the wo rld,

e .g . info rme d(C): c usto me r C is info rme d o n the firm po lic ie s

Ac tion/ e ve nt :

pe rfo rme d b y a n a g e nt o r “inte rna l” in the syste m; ha s dire c t a nd po ssib ly indire c t e ffe c ts o n flue nts, c a using so me sta te c ha ng e (unle ss e ffe c t a lre a dy true )

Ac tion laws: dire c t e ffe c ts o f a c tio ns Causal laws: flue nt de pe nde nc ie s, a nd the n indire c t

e ffe c ts o f a c tio ns

Pr e c ondition laws: a c tio n c a n ha ppe n o nly if

pre c o nditio ns ho ld

T e mpo ra l mo da litie s in DL T L

A pr

• gr

am is b uilt fro m a c tio ns using “+” (o r), “;”

(se q ue nc e ) a nd “*” (ite ra tio n): a | π1 + π2 | π1 ; π2 | π * T e mpo ra l fo rmula e inc lude :

〈π 〉 α

the re is a n e xe c utio n o f π a fte r whic h α ho lds [π ] α

α ho lds a fte r a ll po ssib le e xe c utio ns o f π

[a] α

α ho lds a fte r a

a nd the usua l te mpo ra l lo g ic mo da litie s: ◊α (e ve ntua lly α), □α (a lwa ys α), ○α (ne xt α)

(the ir se ma ntic s is de fine d fro m the o ne o f α 𝓥π β whic h me a ns: the re is a n e xe c utio n o f π a fte r whic h β ho lds, a nd α ho lds in a ll pre vio us sta te s)

Ac tio n la ws

Ge ne ra l fo rm (l

0 flue nt lite ra l – a flue nt o r its ne g a tio n,

l

1 , ... , l n flue nt lite ra ls o r te mpo ra l lite ra ls [a] l ):

□ ([a] l

0 ← l 1 , ... , l m , no t l m+1 , ..., no t l n)

me a ns tha t: a lwa ys, if a is e xe c ute d in a sta te whe re l

1 , ... , l m

ho ld, a nd the re is no re a so n fo r l

m+1 , ..., l n to ho ld,

the n l

0 ho lds in the re sulting sta te

E xa mple :

□ ([info rm]info rme d)

Pe r siste nc y laws:

□ ([a] l ← l , no t [a] ¬ l )

Ac tio n la ws

No nde te rministic a c tio n la ws:

□ ([a] (l

0 ∨ ... ∨ l k ) ← l k+1 , ... , l m , no t l m+1 , ..., no t l n)

i.e ., o ne o f l

0 ,... , l k b e c o me s true .

T he y a re no t primitive , c a n b e ma ppe d to a se t o f a c tio n la ws using de fa ult ne g a tio n

Ca usa l la ws

Ge ne ra l fo rm (fo r static c a usa l la ws):

□ (l ← l

1 , ... , l m , no t l m+1 , ... , no t l n )

mo de l de pe nde nc ie s a nd indire c t e ffe c ts: if l

1 , ... , l m

a lre a dy ho ld o r a re c a use d to ho ld, a nd l

m+1 , ... , l n do

no t ho ld o r a re c a use d no t to ho ld, the n “l” a lso ho lds (if its c o mple me nt ho lds b e fo re , it do e s no t pe rsist) E xa mple :

□ (¬ c o nfirme d ← de le te d)

whe re “c o nfirme d” me a ns “o rde r c o nfirme d fo r the se lle r” – b ut the c usto me r c a n withdra w its o rde r, ma king “de le te d” true a s a dire c t e ffe c t, a nd “c o nfirme d” fa lse a s a side e ffe c t

Ca usa l la ws

Dynamic c a usa l la ws:

□ (О l ← t1 , ... , tm , no t tm+1 , ... , no t tn )

T he ti ’ s c a n b e o f the fo rms l

i o r О l i

T he n we c a n re pre se nt side e ffe c ts o f c hang e s o f flue nts, e .g .:

¬ f , О f

i.e . f b e c o me s true

Ra mific a tio ns & BPs

[We b e r e t a l. 2010] pro po se to use c la use s (in c la ssic a l lo g ic ) to mo de l de pe nde nc ie s, a nd the Po ssib le Mo de ls Appro a c h [Winsle tt 1988] to de a l with ra mific a tio ns the inte nde d sta te s a fte r a n a c tio n a re tho se :

• whe re dire c t e ffe c ts ho ld
• whe re the b a c kg ro und a xio ms a re sa tisfie d
• tha t diffe r minima lly fro m the sta te b e fo re the a c tio n

But o ne o f the ir e xa mple s is: insura nc e c la im a c c e pte d whe n a c c e pte d b y re vie we r A a nd b y re vie we r B

Ra mific a tio ns & BPs

I f this is mo de le d a s the ma te ria l implic a tio n: c laimAc c Re vA ∧ c laimAc c Re vB ⊃ c laimAc c e pte d a nd the PMA is use d, if A a lre a dy a c c e pte d a nd B a c c e pts, this e ithe r ma ke s c laimAc c e pte d true ... o r c laimAc c Re vA fa lse !!! T he sta tic c a usa l rule c laimAc c e pte d ← c laimAc c Re vA , c laimAc c Re vB c a n b e use d to ha ve o nly c laimAc c e pte d c ha ng e a s a side e ffe c t, while still inte nding tha t the implic a tio n ho lds

Ra mific a tio ns & BPs

T he implic a tio n ma y b e fa lse if e .g . we a llo w the a c c e pta nc e to b e o ve rridde n la te r b y a supe rviso r I n this c a se dyna mic la ws a re a ppro pria te :

О c laimAc c e pte d ← О c laimAc c Re vA , ¬ c laimAc c Re vB, О c laimAc c Re vB

i.e ., if the c o njunc tio n o f a c c e pta nc e s b e c o me s true , we ha ve the side e ffe c t, whic h:

• re ma ins true b y de fa ult pe rsiste nc e
• ma y b e ma de fa lse while its o rig ina l c a use re ma ins true

Mo de ling Busine ss Pro c e sse s

A do ma in de sc riptio n in [Gio rda no e t a l T PL P 2012] is a pa ir (Ac tio n a nd c a usa l la ws, DL T L c o nstra ints) T he c o ntro l flo w o f a b usine ss pro c e ss c a n b e mo de le d in se ve ra l wa ys (no t mutua lly e xc lusive )

Option 1: a pro g ra m (re g ula r e xpre ssio n) in a DL

T L c o nstra int: 〈π〉 ⊤ (o nly struc ture d, se q ue ntia l pro g ra ms)

Option 2: use g e ne ra l DL

T L c o nstra ints, e .g .

□ [a ] 〈 b 〉 ⊤ □ [a ] ◊ 〈 b 〉 ⊤

a fte r a is e xe c ute d, imme dia te ly/ e ve ntua lly b is e xe c ute d

Mo de ling Busine ss Pro c e sse s

Option 3: use Co nDe c c o nstra ints (g ive n the ir L

T L c o rre spo nde nt)

Option 4: use e ffe c ts o f so me a c tio ns a s pre c o nditio ns o f

• the r a c tio ns

Option 5: use a «c la ssic a l» g ra phic a l wo rkflo w no ta tio n

(BPMN, YAWL ) a nd de fine a tra nsla tio n to te mpo ra l ASP Ac tua lly, g ive n tha t we tra nsla te te mpo ra l ASP to pla in ASP, we de fine d a dire c t tra nsla tio n fro m a YAWL sub se t to ASP

Da ta

Ac tio ns tha t a c q uire a va lue fo r a va ria b le in a finite do ma in: no nde te rministic a c tio n la ws [ve rify_status] status(g o ld) ∨ status(silve r) ∨ status(no rmal) F

• r nume ric a l da ta , we c a n use the a b stra c tio n

te c hniq ue in [K nuple sc h e t a l 2010] whic h use s thre sho lds in the mo de l (XOR-splits) a nd in the fo rmula to b e ve rifie d to re duc e the do ma in to a sma ll se t o f a b stra c t va lue s

Da ta

We use the no ta tio n (fro m so me ASP so lve rs) 1{[a ]R(X) | P(X)}1 a fte r a , R is true fo r e xa c tly o ne X suc h tha t P(X), sho rt fo r: [a ]R(X) ← no t [a ] ¬ R(X) , P(X) [a ] ¬ R(X) ← [a ]R(Y ) , P(X) , P(Y ) , X ≠Y T he n «se le c t a shippe r S, a mo ng the a va ila b le o ne s, tha t is c o mpa tib le with the pro duc t P» is: 1{[se le c t_shippe r(P)]shippe r(S) | a va ila b le _shippe r(S)}1

⊥ ← [se le c t shippe r(P)]shippe r(S) , no t c o mpa tib le (P, S)

Co mplia nc e

Se ve ra l no rms a nd b usine ss rule s ha ve the fo rm «if A ha ppe ns/ is true , tha n B sha ll ha ppe n/ b e true » T his ma y me a n ve rifying the fo rmula :

□ (A → ◊ B)

b ut this do e s no t a llo w the «o b lig a tio n» to B to b e c a nc e lle d la te r: e .g . (o b lig a tio n to se nd g o o ds c a nc e lle d if o rde r c a nc e lle d) We the n ha ve a n e xplic it no tio n o f c o mmitme nt C(α):

□ ([o rde r]C(g o o ds_se nt))

Co mplia nc e

T he c o mmitme nt ma y b e c a nc e le d:

□ ([c anc e l_o rde r] ¬ C(g o o ds_se nt))

a nd it is disc ha rg e d whe n fulfille d:

□ (○ ¬ C(α)← C(α) , ○α)

(dyna mic c a usa l rule ) No w, the fo rmula to b e ve rifie d is, fo r a ll C(α) :

□ (C(α) → ◊ ¬ C(α))

Co mplia nc e

Othe r b usine ss rule s (no t re la te d to c o mmitme nts) c a n b e ve rifie d, e .g .: ”Pre mium c usto me r sta tus sha ll o nly b e o ffe re d a fte r a prio r so lve nc y c he c k” □(so lve nc y c he c k do ne ∨ ¬ 〈o ffe r pre mium status〉 ⊤)

Co mplia nc e

Ve rific a tio n is pe rfo rme d with Bo unde d Mo de l Che c king in Answe r Se t Pro g ra mming [Gio rda no e t a l, T PL P 2012, K R 2012] T he T e mpo ra l ASP re pre se nta tio n is tra nsla te d to sta nda rd ASP, a nd BMC is re pre se nte d in ASP (fo llo wing [He lja nko & Nie me lä 2003]) I n pa rtic ula r, we use d the ASP syste ms DL V a nd Cling o

Co nc lusio ns

We sho we d so me c o ntrib utio ns fro m Re a so ning a b o ut a c tio ns a nd c ha ng e in AI a nd Answe r Se t Pro g ra mming to Busine ss Pro c e ss mo de ling a nd ve rific a tio n:

• Mo de ling dire c t e ffe c ts o f a c tivitie s a s we ll a s

side e ffe c ts with c a usa l rule s

• F

le xib le mo de ling o f o b lig a tio ns

• Mo de ling da ta
• De c la ra tive o r pro c e dura l pro c e ss mo de l
• Ve rific a tio n o f c o mplia nc e with no rms a nd

b usine ss rule s in ASP

T ha nks!