Norms and Commitment for iOrgs(TM) Information Systems: Direct Logic(TM) and Participatory Grounding Checking

Published in ArXiv 0906.2756 November 6, 2010 Page 1 of 18 Norms and Com mitm ent for iOrgs TM Inform ation Syst ems: Direct Logic TM and Participator y Grounding Check ing Carl Hewitt http://carlhewitt.info iOrgs TM Information Systems raise important issues for formalizing norms that require extensions and revisions of previous foundational work. For example, extension and revision is required of t he funda mental assumption of the Event Calc ulus: Time-varying properties hold at particula r time-p oints if th ey have been initiated by an action at some earlier time-po int, and not termin ated by anoth er action in the meantime. The fundamental assumption of the Event Calculus is overly simplistic when it co mes to organizations in which time-varyi ng pro perties have to be a ctively maintained and mana ged in order to continue to hold an d termination b y another action is not req uired for a property t o no longer hold. I .e., if active measures are not taken then things will go haywire by default. Si milarly extension and revision i s req uired for Grounding Chec king pr operties of systems based on a set of gro und i nferences. P reviously Model Checking has been perfor med usi ng t he model o f nondeterministic automata based on s tates determined by time -points. These nondeter ministic auto mata ar e not suitable for iOrgs, which are highl y structured and o perate asynchronously with onl y loosely bounde d nondeter minism. iOrgs I nformation Syste ms have been de veloped as a technolog y i n which organizatio ns have people that are tightly integrated with information technology that enables them to function organizationally . iOrgs formalize existing practices to provide a framework for add ressing issues of authority, accountability, scalability, and rob ustness using methods that are ana logous to human orga nizations. In general  iOrgs are a natural exten sion Web Ser vices, which are the standard for distributed co mputing and software application i nteroperabilit y in large-scale Or ganizational Co mputing.  iOrgs are structured b y Organizational Commit ment that is a spec ial case of Physica l Co mmitment that is defined to be infor mation pledged. iOrgs norms are used to illustrate the follo wing:  Even a very simple micro theory for normative reason ing can engender i nconsistency In practice, it is impossible to verify the co nsistency of a theor y for a practical do main.  Improved Safety in Reaso ning. It is not safe to use clas sical logic and pr obability theory in practical reasoning. November 6, 2010 Page 2 of 18 Contents iOrgs TM Information Systems ....................................................................... 3 Participatory Semantics ................................................................................ 3 Contrast between Participatory Semantics and the Event Calculus ............. 4 Commitment ................................................................................................. 6 Participatory Grounding Checking ............................................................... 7 Direct Inference ............................................................................................ 9 Collusion at Santa Cruz FishMarket ........................................................... 10 Inconsistent policy in War .......................................................................... 10 Paradigm shift from Inconsistency Denial to Rapid Recovery .................. 11 Norms at Santa Cruz FishMarket ............................................................... 11 Conclusion .................................................................................................. 12 Acknowledgments ...................................................................................... 12 Bibliography ............................................................................................... 12 Appendix 1. A simple auction procedure in ActorScript ........................... 15 End Notes ................................................................................................... 16 November 6, 2010 Page 3 of 18 iOrgs TM Inform ation Systems In the organization lie s the power. iOrgs Information Systems have been developed as a technology in which organizations have people that are t ightly int egrated with inform ation technology that ena bles t hem t o function organizat ionally [Hewitt and Inman 1991; Hewitt 2008b, 2008d, 2008g ]. i iOrgs formalize existing practices to provide a f ram ework for addressing issues of authority, accountability, scalability, and robustness using methods that are analogous t o human organiz ations. ii . In general  iOrgs are a natural extension Web Ser vices, which are t he standard for distributed com puting and software appl ication interope rability in large- scale Organizational Computing .  iOrgs are structur ed by Organizationa l Commitment that is a special case of Physical Commitment [Hewi tt 2007 2008b] tha t is defined to be info rmation pled ged . iii This paper discusses how iOrgs require Direct Lo gic in reasoning and participatory grounding checking in syst ems analysis: 1. The development of iOrgs and the extreme dependence of our society on these systems have introduced new phenomena. These systems have pervasive inconsistenc ies among and within the follow ing: o Norms that express h ow systems can be used and te sted in practice o Policies that exp ress over- arching justification for system s and t heir technol ogies o Practices that expres s implem entat ions of sy stems Different parties are responsible fo r constructing , evolv ing, justifying and maintaining practices, norms, and operations for large -scale Organizational Computing. In specific cases any one consideration can trum p t he othe rs. Someti mes debates over inconsistencies c an become quite he ated, e.g., between sales, engineering and financ e. 2. Grounding check ing iv is a fundam ental tool in the ana l ysis of iOrgs. How ever, previous work on m odel checking has been performed us ing the m odel of nondeterm i nistic autom at a based on states d etermined by time-points. These no ndeterministic automata are not suitabl e for iOrgs, which are h ighly structured and oper ate asynchronously with only loosely bounded nondete rminism. Instead an alysis based on r egions of spac e -time (as in Participatory Sem anti cs [Hewit t and Manning 1996 ] ) is required. Participator y Semantics Participatory Semantics [Hewitt and Manning 1996] is base d on regions of space-time called Participations v according to the following leg end: vi November 6, 2010 Page 4 of 18 XML vi i (a message or d ata structure ) A participation A reference A messag e transmission arrived at the spe cified tim e An object transpo r t Contrast bet ween Participator y Semant i cs and the E vent Calculu s Participatory Sem anti cs is based on 4- di mensional regions o f space-tim e whereas the Even t Calculus is based on events (which take 0 tim e) on a g lobal universal tim e -line. This paper does not use the usual Event Calculus formalism [Kowalski and Sergot 1986 , Miller and Shanahan 1999]. A principle reason for not adopting t he Event Calculus is avoidance of its fundamental assum pti on: “ Time-vary ing properties hold at particular time-points if they have bee n initiated by an action at some earlier tim e -point, and not term inated by anothe r action in t he meantim e. ” The fundamental assum ption of the Event Calculus is overly sim plistic when it com es to or ganization s in which time-varying properties h ave to be ac tively mana ged in orde r to continue to ho l d and termination by another action is not required for a property to no longer hold. I.e., if ac tive measures are not taken then things will go haywire by default. For example consider the foll owing property: “ Drive safely” It mig ht be sai d that the property was “terminated” when a driver collides with another vehicle. Howev er , it is often the cas e t hat som e “unsafe driving” occurred before the co llision! XML Participation Source Target Origin Destination Message@Ti me Origin Destination Object Figure 1. Legend November 6, 2010 Page 5 of 18 Another problem with the Event Calculus is that it i s formulated at the ve ry low level of abstraction of time- points. As Carlo Rovell i has explained : vi ii “We never really see time. We see only clocks. If you say t his object moves, what you really mean is that t his object is here when the hand of your clock i s here, and so on. We sa y we measure time with clocks, but we see only the hands of the clocks, not time itself. And the hands of a clock are a physical variable like any other. So in a sense we cheat because what we really observe ar e physical variables as a function of other physical variables, but we represent that as if everything is evolving in time .” Preoccupation with global time-points is a serious problem with the Event Calculus. This problem i s closely r elated t o another problem with the Event Calculus: A t ime i s not bo und to a locale but is instead im agined to be free floa ting! Consider an example in which safe driving is followed by unsafe driving leading to a collision. An important issue is that there may be no ev ent which clearly delineates the transition from safe driving to unsafe driving. ix The lack of such an event is not material t o Participatory Semantics. Howev er, t here is no clear terminating event for t he Event Calculus. Next consider an example in which AM transitions to PM on July 31, 2008 in California. Here the issue is that there is no physical event that occurs throughout California that mark s the transition from A M t o PM. Driv ing Collis ion Unsa fe Dr ivin g Safe D rivi ng Figure 2. Transit ion from Sa f e to Unsafe Dr i ving (with no clipping eve nt) Califor nia on J uly 31, 2 008 PM AM Figure 3. Transit ion from A M t o PM (with no cl i pping event ) November 6, 2010 Page 6 of 18 By conv ention, the AM and PM reg ions for Califo rnia are ad j acent to each oth er. Howev er , this adjacency does not require the existence of any event that occurs t hroughout Californ ia and the lack of such an event is not material to Participatory Semantics. However, there is no terminating event as required by the Event Calc ulus. Commitm ent According to [Hewit t 2007], a Physical Commitment PC is defined to be a pledge that certain information I holds for a physical system PS for a space-time region R . Note that physical commitm ent is defined for whole ph ysical system s ; not just a par ticipant or p rocess. This paper uses a mythical Santa Cruz FishMarket x to illustrate how organizational commitments can be formulated at a higher level of abstraction. Th e Santa Cruz FishMarket uses an electronic English Auction starting with a reserved price in which a certain ti me is allowed for more bids to come in before the bidding is closed. As each higher bid is received, the new minimum bid i s announced to the participants. Tie bids are broken by choosing the one which arrived first as t he winner. Consequently the Santa Cruz Fish Market is an organization al commitm ent with an auction of bu yers and sellers. An implementation xi f or a Sim pleA uc tion for t he Santa Cruz FishMarket is given in the appen dix . xi i The commitm ent s below are m ade by the implem ent ation pledg ing the following information: Figure 4. Commitment: Bidding too late causes an exception Bidde r 1 Simple Auc tion arrivalTime > d eadline Bid.[ amountBid bidder : Bidder 1 ] @ arrivalTim e TooLate . [ deadline ] Exception Request November 6, 2010 Page 7 of 18 Participator y Grounding Check ing The denotationa l s emantics of concur r ency were fi rst developed in [Clinger 1981]. Subse quently [Hewitt 2006b] developed the TimedDiagram s model with the Compu tational Representation Theorem which s tates: The denotation Den ote S of a closed xi ii system S represen ts all the po ssible behavio rs of S as Denote S = ⊔ i  ω Progression S i ( ⊥ S ) where P rogression S is an approximation function t hat t akes a set of approximate behaviors to their next stage and ⊥ S is the initial behav ior of S . The denotational semantics exhibits relatively unbounded nondeterm i nism because i n the deliv ery of a message can occur a relatively unbounded am ount of ti me after it is sent. This relatively unbounded nonde t erm inism can cause trouble with traditional global state-based appr oaches [Bianculli, Morzenti, Pradella, Sa n Pietro, Spoletini 2007; Bordini, Fisher, Visser, and Wooldri dge 2004; Cliffe, De Vos, and Padget 2006; Desai, Cheng, Chopra, and Singh 2007 ; Venkatram an and Singh 1999, etc .] be ca use of th e following iss ues:  State explosion because of the increase in po ssible interleavings of g lobal states  Modeling failure because the system bei ng m odel ed is not fin ite state Participatory grounding checking makes use of the Representation Theorem to characterize pos sible alternative com putations. In parti cipatory g rounding check ing:  Explosion i s less of a problem because local groundings are modeled instead of global state. Also Participatory Semantics can be used t o abstract high level properties of denotations in a way that is sim i lar to how a bstraction has been used in global state m odel checking . Simple Auc tion alarmTime > deadline Alarm .[ alarmTime ] ProcessOutco me Curr entBid ding Request Figure 5. Commitment: An alarm triggers the processing of the auction outcome November 6, 2010 Page 8 of 18  Systems are not modeled as nondeterm ini stic st ate machines, Petri nets, or process cal culi [ Aceto and Gordon 2005] . xi v  Communication i s modeled as being fundamentally one -way and as ynchronous . In this way modeling problems such as occur using Petri Nets and synchronous process calculi are avoided [ Padget an d Bradford 199 8] . For example consider the system with Sim ple Auct ion ( defined i n the appendix of this article) augmented with b idders like the fo l lowing: xv SimpleBidder  behavior {  xvi serialize the messa ges received by this bidder Auction ::theA uction  auction that bidder is bidding for Currency ::m axi mumB id  maximum t hat bidder w ill bid for this auct ion ----- ---- -------- ------- ------- ------ ----- -------- ------- ---- ------ ---- -------- ------- ------ ---- ----- implements Bidder  the Bidder interface i s implemented below self. newMinim um (amount)   a newMin imum message with amount has be en received amount ?? { ( < max imumBid) :  if amount is less t han ma ximu mBid theAu ction . b id (amount for: self)  then b id the minimum catch er {  and if it throws t he exception TooLittle .[ minimumBid ] :  that the bi d was too small relay s elf . newMinimu m (minimumBid) }  send yours elf the new m inimum bid not ( < maxim umBid) : void } }  else do nothing and return vo i d Using bidders like the above, execution scenarios of the system can be computed using the Representation T heorem . These execut ion scenarios ca n be checked against norms such as the following: Commitment: At Santa Cruz FishMarket, deliveries from seller to buyer are paid at the agreed price; i.e. S anta Cruz F ishMarket pledges the followin g information: For every seller , buyer , and deliver y, Delivers [ seller , buyer , delivery ] ├ FishMarket PaysAgr eedPrice [ buyer , seller , delivery ] xvi i Note that sel ler , buyer , and delivery are al l space-time participations in the above norm. Consequently , there is enough information to specify that the buyer pays the agreed price to the sell er on deli very of the purchase. xvi ii November 6, 2010 Page 9 of 18 Figure 6. A commitment involving the exchange of Goods for Payment Direct Inference Direct inference is used in to direct l y infer conclusions f rom prem i ses. For example, suppose that we have 1) WeekdayAt5PM ├ Boston TrafficJam which says that in Bos ton , a we ekday at 5PM infers a traffic jam. 2) ├ Boston  TrafficJam which says that in Bos ton , no tra ffic jam. In classical logic,  WeekdayAt5PM is inferred in Boston fr om 1) and 2) above. But fortunately in Dir ect Logic: ⊬ Boston  WeekdayAt5P M which says in Boston, there is a particular proposition (  WeekdayAt5PM ) that cannot be inferred in Direc t Logic fro m 1) and 2) above . Consequently , direct inference co m es into p lay even in the ab se nce of ove rt inconsistenc y. xix Selle r 1 Buy er 1 Goods Payment Deliv ery 1 November 6, 2010 Page 10 of 18 Collusion at S a nta Cruz FishMark et Ethical conduct at the Fi shMarket is wo rthy of furt her s tudy. A cl assic guide to norms f or ethical conduct appears in The Prince [Machiavel li 1532]. So roughly i n this spirit , cons ider t he follow axiomatization xx of collusion at F ishMarket: xxi 1) p, action: 󰇟󰇟  󰇠󰇟󰇠 󰇠󰆜 ThePrince Do[p, action]  axiom xxii 2) ├ ThePrince CanResult[Do[Mac hiavelli, Collude], R ich[Machiavelli]]  a xiom 2 ’ ) Do[Machiavelli , Collude] ├ ThePrince CanResult[Rich[Machiavell i]]]  axiom 3) ├ ThePrince Do[ Mach ia velli, C ollud e]  from 1) and 2) 4) p, action:  󰇟󰇟 󰇠󰇟󰇠󰇠 󰆜 ThePrince  Do[p, action]  axiom xxiii 5) ├ ThePrince CanResult[Do[Machiav elli, Collude], Ruined[Machiave lli]]  axiom 5’ ) 󰇟 󰇠󰆜 ThePrince CanResult[Ruined[Mac hi avelli]]]  axiom 6) ├ ThePrince  Do [Ma chia velli, C ollud e]  from 4) , and 5) Note that, in t he ThePrince , t here is an inconsistenc y between 3) and 6) . However, even though the ThePrince is i nconsis tent, it is not meaningless because in some respects it captures some underlying paradoxic al aspects of Mac hiavelli‟s theo ry : “ The wish to acquire mo re is admittedly a very nat ural and common th ing; and when men succeed in this they are always praised rather than condemned. But when they lack the abili ty to do so and yet want to acquire m ore at all cos ts, th ey deserve cond emnation for their mistakes.” Inconsistent polic y in War A classic case of inconsistency occurs in the novel Catch-22 [Heller 1961] which states that a person “ would be crazy to fl y more missions and sane if he didn't, but if he was sane he had to fly them. I f he flew them he was crazy and didn't have to; but if he didn't want to he was sane and had to. Yossarian was moved very deeply by the absolute simplicity of this clause of Catch -22 and let out a respectful whist le. „That's some catch, that Catch- 22,‟ he observed. ” In the spirit of C at ch-22, consi der the follow axiomatization of the ab ove: 1. ├ Catch- 22  (Able[x, Fly]   Fly[x])   (Sane[x])  axiom 2. ├ Catch- 22  (Sane[x])   (Obligated[x, Fly])  axiom 3. ├ Catch- 22  (Sane[x]  Obligated[x, Fly])   (Fly[x])  axiom 4. ├ Catch- 22  (Fly[x])   ( Crazy[x])  axiom 5. ├ Catch- 22  (Crazy[x] )   (  Obligated[x , Fly])  axiom 6. ├ Catch- 22  (Sane[p]   Obligate d[p, Fly])   (  Fly[p ])  axiom November 6, 2010 Page 11 of 18 For Yossarian, we hav e t he fol lowing axiom s: 7. ├ Catch- 22  (Able[Yossarian, F ly ])  1  axiom 8. ├ Catch- 22  (Sane[Yossa rian ])  1  axiom Consequently , 2’. ├ Catch- 22 1   (Obligat ed[Yossarian, Fly])  Yossarian using 2 and 8 3’. ├ Catch- 22 1   (Fly[Yo s sarian ])  Yossa rian using 3, 2’ and 8 4’. ├ Catch- 22 1   (Crazy [Yossarian ])  Yossarian us ing 4 and 3’ 5’. ├ Catch- 22 1  1 -  (Obligated[Yossarian , Fly])  Yossarian using 5 and 4’ 5’’. ├ Catch- 22  (Obligated[Yossarian, Fly])  0  reformulation o f 5’ 6’. ├ Catch- 22 1  1 -  (Fly[Yossarian ])  Yossarian using 6 ’ , 8 and 5’’ 6’’. ├ Catch- 22  (Fly[Yossar ian ])  0  reformulation o f 6’ Thus there is an inconsisten cy in Catch- 22 in that both of the following hold: 3’. ├ Catch- 22 1   (Fly[Yo s sarian ]) 6’’. ├ Catch- 22  (Fly[Yossar ian ])  0 Paradigm shift from Inconsist ency Denial to Rap id Recover y ThePrince and Catch- 22 illustrate the following i mportant points:  Even a ve ry simple m icrotheory for norma t ive reasoning can engender inconsistency In practice, it is impossible to verify the consistency of a t heory for a prac tical dom ain.  Improved Safety in Re asoning . It is not s afe to use classica l logic and prob ability theory in practical reason ing. Norms at Santa Cru z FishMark et Norms are com mit ment sup ported by comm uniti es of p ractice. Norm: At Santa Cruz FishMarket, there is no collusion among buyers and sellers. Formalizing the no rm above is the sub j ect of future r esearch. November 6, 2010 Page 12 of 18 Conclusion iOrgs raise important issues f or inconsistency robustness and participato r y grounding checking. This paper presents some ideas for formalizing these issues. Relationships among these issues are analyz ed using illustration s from FishMarket and Th e Prince . The following conclusions ar e proposed.  Extension and rev i sion is re quired of the fun damental a ssumption of the Ev ent Calculus: Time-varying prope rties hold at particular tim e-points if they have bee n i nitia ted by an action at some ear lier time-poin t, and not term inated by another a ction in the meantime. The fundamental assum pti on of the Ev ent Calculu s is overly si mplistic when it comes to iOrgs in which time-v arying properties have to be actively main t ained and managed in order to continue to hold and ter min ation by another action is n ot required fo r a property to no longer hold. I.e., if ac tive measures are not taken then things will go hay wire by default. Consequently the Event Calculus a pproach m ust evolve into a strongly paraconsisten t system structured around p articipations in space-time.  Similarly extensio n and revision i s required fo r Model Check i ng properties o f system s. Previously Mode l Checking has been performed usi ng the model o f nondeterm i nistic automata based on s t ates de termined by tim e -points. These nonde terministic auto mata are not suitable for iOrg s, which are highly structured and operate asynch r onously w ith only loose ly bounded nondete rminism. Conseque ntly Model Checking needs to evolve in the di rection of verifying participa tory beha vior in iOrg s. Ackno wledgments The development of Participatory Semantics was joint work with Carl Manning. Les Gasser, Mike Huhns, Victor Lessor, Pablo No riega, Sascha Ossow ski, Jaime Sichman, Munindar Singh, etc. provided valuable sugg est ion s at AAMAS‟07. The reviewers and participa nts of MALLOW‟07 (including J ohn Lloyd, John-Jules Meyer, Pablo Noriega, Jaime Sic hman, Muninda r Singh, Rineke Verbrugge, etc. ) provided valuable comments. Aft erw ards Munindar Singh provided helpful pointers to the literature. T he reviewers f or COIN@A AMAS‟08 made hel pful suggestion s. Conversations with Jeremy Fort h were helpfu l in developing the com parison of Pa r ticipatory Semantics and the Event Calculus. Munindar Singh made ext ensive comm ents and suggestions that significantly i mprov ed the presentation. A helpful review w as prov i ded by an anonym ous referee for the s pecial is sue of the Journal of I GPL on Normative Multi-agent Sy stems. Bibliography  Chris Ander son. The End of The ory: The Data Deluge Makes the Scientific Method Obsolete Wired. June 23, 2009.  Luca Aceto and Andr ew D. Gordon (Ed.) Algebraic Process Calculi: The Fi rs t Twenty Fiv e Years and Beyond 200 5  Marco A lberti, Marco Gavanelli, Evelina Lamm a, Paola Mello, and Paolo T orroni “ Modeling interactions using social integrity constraints: A resource sharing ca se st udy” DALT . LNAI 2990. Springer-Verlag, 2004.  Aldo Antonelli . “Non - monoton i c Logic” Stanfo rd Encyclopedia of Philosophy . March 2006. November 6, 2010 Page 13 of 18  Domenico Biancul l i, Angelo Morzen t i, Matteo Pr adella, Pierluig i San Pietro, Paol a Spoletini “Trio2Promela : A Model Check er for Tem poral Metric Speci fications” ICSE'07 .  Brian Blum. C ontracts: Exam ples and Explana tions Aspen Publishe rs. 3 rd edition 2004.  Rafael Bordini, M ichael Fis her , Willem Visser, and Mi chael Woold ridge “ Verifiable Mul ti-Agen t Programs” Spring er LNAI 3067. 2004.  Geof Bowker, Sus an L. Star, W. Turner, and Les G asser. (Eds.) Social Science Researc h, Technical Systems and Cooperativ e Work . Lawrence E ar lbaum . 1997.  Cristiano Castelfranchi “Comm it ments: From i ndividual intentions to groups and organizations” AAAI -93 Workshop on AI and Theories of Groups and Organizations: Conceptual and Empirical Research .  Amit Chopra and Munindar Sing h. Multiagent Comm itment Alignm ent . Draft of Ma rch 7, 2009.  Owen Cliffe, Mar i na De V os, and Julian Padget (2006). “Specifying and Reasoning about Multiple Institu tions” COIN@ AAMAS‟06.  Jiangbo Dang, J ingshan Huang and Mike Huhns “Workflow Coordination for Service -Oriented Multi- agent Sy st ems” AAMAS‟07 .  Nirmit Desai, Zhengang Cheng, A mit Chopra, and Munindar Singh “Towar d Verification of Comm itment Protocols and their Composi tions” AAMAS‟07 .  Virginia Dignum A Model for Organizational In t eraction. PhD thesi s. Utrecht. 2004..  Rogier Van Eijk, Frank De Boer, Wieb e Van Der Hoek and J ohn-Jules Meyer . “A verification framework for agen t comm unication” AAMAS ‟03 .  Andrew Farrell, Marek Sergot, Mathias Salle and Claudio Bartolini “Using the event ca lculus f or tracking the normative state of contracts” International Journal of Cooperat ive Information System June-Septem ber 2005.  Tim Folger . “New sflash: Ti me May Not Exist” Discover. J une 12, 2007.  Dorian Gaertner, Pablo Noriega, J. -A. Rodriguez-Aguilar, Wamberto Vasconcelos (2007). “Distributed Norm Ma nagem ent in Regul ated Multi- Agent System s” AAMAS‟07.  Joseph Heller. Cat ch-22 Sim on and Schus ter. 1961.  Carl Hewitt and J eff Inm an (1991). “DAI Betwixt and B etween: From „ Intelligent Agents‟ to Open System s Science” IEEE Transactions on Sy st ems, Man, and C ybernet ics. Nov. /Dec. 1991.  Carl Hewitt and C arl Manning ( 1996). Synthetic In frastructures f or Multi-Agency Systems ICMAS'96.  Carl Hewitt W hat is Commitment? Phys i cal, Organi zational, and Soci al (Revised) Pablo No riega .et. al. editors. LN AI 4386. Springer-Verlag. 2007.  Carl Hewitt ( 2008a) Large-scale Organizational Computing requires Unstratified Reflection and Strong Paraconsistency Coordination, Organizations, Institutions, and Norm s i n Agent Systems III. J aime Sichman, Pablo Noriega, J ulian Padg et and Sascha Ossowski (Ed.). Springer- Verlag. 2008. http://organi zational.carlhewitt.info/  Carl Hewi t t (200 8b) . O RGs f or Scalable, Robust, Privacy-Friendly C l ient Cloud Com puting IEEE Internet Com puting September/October 2008.  Carl Hewitt ( 2008c) Middle History of Log ic Pro gramming: Resolution, Planner, Edinburgh Logic for Computable Functions, Prolog and the Japanese Fifth Gen eration Project ArXiv 0904.3036  Carl Hewitt (2008d) A historical perspective on developing foundations for client cloud computing: iCon sult TM & iEntertain TM Apps using iInfo TM Information Integration for iOrgs TM Information Systems (Revised version of “Dev elopm ent of Log i c Programm ing: W hat went wrong, What was done about it, and What it might mean for the future” AAAI Workshop on What Went Wrong. AAA I-08.) ArXiv 090 1.4934.  Carl Hewitt (200 8e) Common se nse for concurrency and inconsistenc y tolerance u sing Direct Logic TM and the Actor Mod el ArXiv 0812.485 2. November 6, 2010 Page 14 of 18  Carl Hewitt (2008f) Scalable Privacy-Friendly Client Cloud Computing: a gathering Perfect Disruption Stanford C omputer Sy st ems Laborato r y Colloquium Oct. 22, 2008. Video record i ng. http://stanford-on line.stanford.edu /courses/ee380 /081022-ee380-300.asx  Carl Hewitt (2008g ) Perfect Disruption : Causing the P aradigm Shift from Mental Agents to ORGs IEEE I nter net Com puting. Jan/Feb 2009.  Carl Hewitt ActorS cript™: Industrial streng th integrat ion of local and no nlocal concurrency fo r Clie nt -cloud Computing ArXiv. 0907. 3330  Bryan Horling and Victor L esser (2005). “Us i ng OD ML to Model Multi - Ag ent Organiz ations” IAT‟05  Nicholas Jenning s “Comm itments and conv entions: The foun dation of coord i natio n in multi - agent systems” Know ledge Engine ering Review. 3. 199 3.  Frederick Knabe “ A D istributed P rotocol for Chan nel - Based Comm unica tion with C hoice” PARLE’92  Bill Kornfeld and C arl Hewitt “The Scientific Com munity Metaphor” IEEE Transa ctions on Systems, Man, an d Cybernetics. January 1981.  Robert Kowals ki and Marek Sergot. “A Logic - based Calculus of Ev ent s” New Genera t ion Computing Volum e 4 , Issue 1. 1986.  Robert Kowalski. “Database upda tes in the Event Calculus” Journal of Logic Programm i ng . 1992.  Niccolò Machiav elli (1532) The Prince (Bantam Clas sics 1984 ).  Robin Milner “Elem ents of interaction: Turing award lectu re.” CACM. January 1993.  Pablo Norieg a Agent Mediated Au ctions: The F ishmarket Metaphor . Ph.D. Barcel ona. 1997.  Julian Padget and Ru ss ell Bradford (1998 ). “A π -calculus Mode l of a Sp anish Fi sh Market” AMET-98 also in LN AI 1571 Spring er Verlag. 1999.  Carl Petri Kommun i kation m i t. Automate . Ph. D. Thesis. University of Bonn. 1962 .  Murray Shanahan and Rob Miller (1999) “The ev ent -calculus in cla ssical logic — alternative axiomatization, ” Electronic Transactions on Ar t ificial Intelligence 3(1).  Jaime Sichman, Virg inia Di gnum, and Cris tiano Castel franchi (2005). “Agent organizations.” JBCS , 11(3).  Munindar Singh. “So ci al and Psy chological Com mitments in Multi - agent System s” AAAI Fall Symposium on Knowl edge and Action at Social an d Organizational Le ve ls. 1991  Munindar Singh “An ontology for commitments in multi -agent systems: Toward a unification of normative conc epts” Artificial Intelligence and Law 7. 1999.  Feng Wan and Munindar Singh (2005). “Formalizing and achieving multiparty agreements via commitm ent s” AAMAS‟05 .  Michael Winikoff, Wei Liu, and James H arland (2005). “Enhancing comm itment machines” DALT . LNAI 3476. Spri nger-Verlag . 2005.  Pinar Yolum and Mu ni ndar Singh “Enac t ing Protocols by Commitm ent Concession” AAMAS‟07 .  Mahadevan Venkatraman and Munindar Singh. “Verifying Compliance with Commitment Protocols: Enabling Open Web-Based Multi- agent Systems” JAAMAS. 1999. November 6, 2010 Page 15 of 18 Appendix 1. A s imple auction procedure in ActorScript SimpleAuction  behavior { Bidders ::theBidders  a collection of those allowed to bid on this auction Currency :: mini mumBid  current minimum bid fo r this auction Time :: dea dline  current dea dline by which this auctio n will end unless  another higher bid is received for this auctio n Bidding :: c urrentBidding  a reco rding of the current state of b idding for this auction ----- ----- ----- ----- ----- ---- - ---- ------- ---------- ---- - ---- - ---- ---------- ----- ----- ----- ----- - implements Auction  the Auction interface is implemented belo w self . b id ( amountB id ) @ T ime :: arrivalTime   a messag e with amount bid an d bidder has been received at arrivalTime arrivalTime ?? {  if arrival Time ( > deadline) : throw TooLate . [ deadline ] ;  is after deadline , co mplain b id is too late not ( > deadline) : amountBid ?? {  else if the amount bid is smaller than the minimum ( < minimumBid) : throw TooLittle . [ minimumBid ] ;  then compla in the bid is too little not ( < minimumBid) : {currentBiddi ng. b id (amountB id time : arrivalTime) ;  record the bid in currentBidding  this may throw an exceptio n if the bidder is un qualified let ( Time :: newDeadline = C urrentTime ( )+  , Amount :: newMinimumBid = amountBid * 110 %)  compute the new dead line and new minimum bid { theBidders . newMinimum (newMinimumB id deadline : newDeadline ),  inform the allowed b idders of the new minimum an d deadline s elf. alarm (newDeadline ) ,  set an alarm for this au ction with a new dead line Acknowledgment . [ ] also become (minimumBid = newMinimumBid , deadline = newDeadline) }} }  return acknowledgment that the bid ha s been accepted and  also update this a uction with the new minimu m bid and deadlin e alarm. (alarmT ime )   an a larm message with al armTime has been received alarmTime ?? {  if ala rmTime ( < deadline) : void;  is before the deadline return void not ( < deadline) : currentBiddi ng . processOutcome } }  else return th e result of p rocessing the o utcome of this auction a ccording to the currentBid ding November 6, 2010 Page 16 of 18 End Notes i The ar chitecture o f an iOrg differs fu ndamentally fro m a Menta l Age nt that cog nitively operates in a hu man-like fashion. The Mental Agent p aradigm [Alberti, Gavanelli, Lam ma, Mello, an d T orroni 2 004] has had som e su ccess in modeling and si mulating hu man-like behavior. Ho wever, co mputing has changed d ramatically fr om the ti me of its in vention a nd we are in the midst of a “perfect disruption” [He witt 2008g] brought on by the following:  Hardware . Many-core architecture th at will soon s upport thousands o f threads in a process for w idely- used software ap plications using semanti c integration (see belo w).  Software . Client cloud co mputing in which infor mation is per manently stored in servers on the Internet and cac hed temporaril y on clients that range from si ngle chip sensors, handhelds, notebo oks, desktops, and entertai nment centers to huge data centers. (Even data centers are clients that often cac he their information to guard against geographical disa ster.) Client cloud co mputing will provide much needed ne w capabilities including t he following: o maintaining t he privacy of client infor mation by storing it on ser vers encrypted so that it can be dec rypted only by using the clie nt‟s private key. (The information is unencr ypted only when cached o n clients.) o providing greater integration of user in formation obtained fro m servers of competing vendors without requiring them to i nteract with each other. o providing better ad vertising relevance and tar geting without exposin g client privacy.  Applications . Scalable semantic inte gration, e.g., integrating the following: o calendars and to do lists o email archive s o presence infor mation including physical, p sychological and social o docu ments (including presentations, spread sheets, propo sals, job applications, photos, videos, gift lists, memos, purchasing, con tracts, articles, etc .) o contacts (includi ng social grap hs) o search results o marketing and ad vertising relevance in fluenced by the above This perfect d isruption is causing a paradigm s hift fro m Mental A gents to i Orgs as the foundation for i mplementing large -scale Internet app lications [Hewitt 200 8g]. ii Fo r background i nformation on iOr gs see [B owker, Star, Turner, and Gasser 1977; Dignum 2004, Singh and Huhns 200 5; Horling and Lesser 20 05]. iii In some previous work, the subject of contracts [Blum 20 04, etc .] has been treated using the (somewhat unfortunate nam e) “commitment” for contractual obli gations [ Singh 19 91, Jennings 199 3; Noriega 1997; Singh and Huh ns 2005; Chopra and Singh 20 09]. In this paper , these obligations are treated as special cases of P hysical Commitment. See He witt [2007] . iv i.e ., checking the beha vior of a system against a model v Note that Particip ations (being regions of space -time) r epresent both objec ts and activities. November 6, 2010 Page 17 of 18 vi Note t hat Participator y Semantics is based on space -ti me as op posed to the more usual approach of basing se mantics j ust on time, e. g. the Event Calculus [Farr ell, Sergot, Salle, and Bartolini 2005 ], etc. vii XML is used be b ecause it is i ncreasingly do minant as the d e facto standard for str uctured message communicatio n and stands to beco me the de fa cto standard for d ata structures. viii quoted in [Folger 20 07] ix For example, such an eve nt would have to exist in the Eve nt Calculus for mulation in Kowalski [1992]. x Inspired by the Blanes Fis hMarket Metaphor [ Noriega 1997]. xi appendix in the concurre nt programming language ActorScript [ Hewitt 2008c] xi i A functional notation is used for XML. For example Perso nName< Firs t<  Kurt  > Last<  Gödel  >> can print as : Kurt Gödel xi ii Closed means that the system does not r eceive any communicatio ns from outside itself. xiv The Acto r Model subsumes other models o f co ncurrenc y, e.g. Process Calculi [Milner 19 93; Aceto and Go rdon 2005] and Pet ri Nets [ Petri 1962] using a t wo -phase commit pro tocol [Knabe 1992 ]. xv Implemented in ActorScript TM [Hewitt 2010] xvi The symbol  is used to begin a co mment that extends to the end of the line. xvi i Expressed in Direct Logic [He witt 2008c] (see discussion la ter in this paper). xviii Note that ( unlike Venkatra man and Si ngh [1999 ]), no assumption is made that t he buyers a nd sellers are not malicious, e.g ., no use is made of time-sta mps that can be forged. xix Statistical probabilistic (f uzzy logic) systems are affected follows: Suppo se (as above) 󰆜 Boston  (TrafficJam | WeekdayA t5PM) = 1 xix 󰆜 Boston  (TrafficJam) = 0 Then 󰆜 Boston  (WeekdayAt5PM) =  󰇛    󰇜   󰇛   󰇜 = 0 xix Thus contrapo sition i s built into probabilistic (fuzzy logic) systems a nd conseq uently i ncorrect information can be generated . The abo ve example ill ustrates t hat t he choice o f ho w to incorpo rate measurements into statistics ca n effectively determine the model being used. In this p articular case, the way that measurements were take n did not happen to ta ke into ac count things li ke holida ys and severe winter storms This point was largely missed in [ Anderson 2008 ] which stated November 6, 2010 Page 18 of 18 “Correlation is enou gh.” W e can stop looking for models. We can analyze the data without hypothese s abo ut what it m ight show. We can throw the n umbers into the big gest computing clusters the world has ever seen and let statistica l algorithms find pattern s w here science cann ot.” (emphasis added ) xx The axio matization makes use of higher order capabilities. For exa mple a predicate like CanResult can ta ke arguments Do[Ma chiavelli, Collude] and Rich[Machiavelli] to form the proposition CanResult[D o[Machia velli, Col lude], Rich[Machiavell i]]. xxi The axio matization uses a colon ( : ) to separate universally quantified variable fro m the following proposition to which they app ly. Direct Logic supports fine grai ned reasoning because i nference does not nece ssarily ca rry argument in the contrapositive directio n . For example, the general principle “A per son does anything that can make them rich” ( i.e., Ca nResult[Do[p, action], Rich[p]] ├ ThePrince Do[p, action] does not suppo rt the inference of  CanResult [Do[Machiavelli, Coll ude], Rich[Machiavelli]] fro m  Do[Machia velli, Collude]). E.g., it might be the case that CanResult[Do[Machia velli, Collude], Rich[Mac hiavelli]] even thoug h it infers Do[Machiavelli, Coll ude] contrad icting  Do[Machiavell i, Collude]. xxii “ A prince never lacks leg itimate reason s to break his pro mise ” [Machia velli 1532] xxiii “ The on e who adapts his policy to the times prospers, and likewise tha t th e o ne whose policy clashes with the dema nds of the times does n ot.” [Machiavelli 153 2]