%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% %% The following file is a template, including explanations, for you to use %% in preparing your group's annual progress report. The progress.sty file %% is available from HQ. %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% \documentstyle[12pt,progress,psfig]{article} \bookfalse %comment \drafttrue out for final run. %\drafttrue %% Don't use this --- draft defintions omitted % Reset MARGINS \setlength{\textwidth}{6.5in} \setlength{\oddsidemargin}{0in} \setlength{\evensidemargin}{0in} \setlength{\topmargin}{-18pt} \setlength{\headsep}{12pt} \setlength{\textheight}{9in} % Floating text boxes to interest the reader \def\floater#1{ \begin{figure}[hbt] \hbox{\hspace{1.5in} % \framebox{\relax \begin{minipage}{5in} \begin{flushright} \large\sl#1 \end{flushright} \end{minipage}} % } \end{figure} } \newlength{\boxwidth} \newcommand{\boxfigure}[1]{\relax \setlength{\boxwidth}{\textwidth} \addtolength{\boxwidth}{-1pc} \begin{figure}[tp] \framebox{\relax \begin{minipage}{\boxwidth}\parindent=1em #1 \end{minipage}} \end{figure}} \makeatletter % not needed in .sty files % mpsg-article.sty: Patches to Latex article.sty by CarlManning % Use by including as \documentstyle{mpsg-article} \pagestyle{headings} % SECTIONS: % don't skip such a large space between number and title \def\@sect#1#2#3#4#5#6[#7]#8{\ifnum #2>\c@secnumdepth \def\@svsec{}\else \refstepcounter{#1}\edef\@svsec{\csname the#1\endcsname\hskip 1ex}\fi \@tempskipa #5\relax \ifdim \@tempskipa>\z@ \begingroup #6\relax \@hangfrom{\hskip #3\relax\@svsec}{\interlinepenalty \@M #8\par} \endgroup \csname #1mark\endcsname{#7}\addcontentsline {toc}{#1}{\ifnum #2>\c@secnumdepth \else \protect\numberline{\csname the#1\endcsname}\fi #7}\else \def\@svsechd{#6\hskip #3\@svsec #8\csname #1mark\endcsname {#7}\addcontentsline {toc}{#1}{\ifnum #2>\c@secnumdepth \else \protect\numberline{\csname the#1\endcsname}\fi #7}}\fi \@xsect{#5}} % % Numbering sections: 1. Title rather than: 1 Title % \def\thesection {\arabic{section}.} \def\thesubsection {\thesection\arabic{subsection}} % % Skip 1ex rather than 1em between number and title % \if@twoside \def\ps@headings{\let\@mkboth\markboth \def\@oddfoot{}\def\@evenfoot{}\def\@evenhead{\rm \thepage\hfil \sl \leftmark}\def\@oddhead{\hbox{}\sl \rightmark \hfil \rm\thepage}\def\sectionmark##1{\markboth {\uppercase{\ifnum \c@secnumdepth >\z@ \thesection\hskip 1ex\relax \fi ##1}}{}}\def\subsectionmark##1{\markright {\ifnum \c@secnumdepth >\@ne \thesubsection\hskip 1ex\relax \fi ##1}}} \else \def\ps@headings{\let\@mkboth\markboth \def\@oddfoot{}\def\@evenfoot{}\def\@oddhead{\hbox {}\sl \rightmark \hfil \rm\thepage}\def\sectionmark##1{\markright {\uppercase{\ifnum \c@secnumdepth >\z@ \thesection\hskip 1ex\relax \fi ##1}}}} \fi \def\ps@myheadings{\let\@mkboth\@gobbletwo \def\@oddhead{\hbox{}\sl\rightmark \hfil \rm\thepage}\def\@oddfoot{}\def\@evenhead{\rm \thepage\hfil\sl\leftmark\hbox {}}\def\@evenfoot{}\def\sectionmark##1{}\def\subsectionmark##1{}} % ITEMIZE: % shouldn't use dash at second level \renewcommand{\labelitemii}{\mbox{$\circ$}} % A version of itemize without space between items, looks better if % items are single lines. \def\itemizeNoSpace{\ifnum \@itemdepth >3 \@toodeep\else \advance\@itemdepth \@ne \edef\@itemitem{labelitem\romannumeral\the\@itemdepth}% \list{\csname\@itemitem\endcsname}{ \itemsep=0pt \topsep=0pt \listparindent=\parindent \parsep=0pt \def\makelabel##1{\hss\llap{##1}}}\fi} \let\enditemizeNoSpace =\endlist % BIBLIOGRAPHY & CITATIONS: \def\secbibliography#1{ \section*{Bibliography}\markright{Bibliography} \list {[\arabic{enumi}]}{\settowidth\labelwidth{[#1]}\leftmargin\labelwidth \advance\leftmargin\labelsep \usecounter{enumi}} \def\newblock{\hskip .11em plus .33em minus -.07em} \sloppy \sfcode`\.=1000\relax\small} \def\endsecbibliography{\endlist\vfill} \def\@biblabel#1{[#1]} \def\@cite#1#2{[{#1\if@tempswa , #2\fi}]} \def\@lbibitem[#1]#2{\item[\@biblabel{#1}]\if@filesw { \def\protect##1{\string ##1\space}\immediate \write\@auxout{\string\bibcite{#2}{#1}}}\fi\ignorespaces} % Adjust \cite for better handling of long citation styles, such as % [Manning, Inman, & Hewitt 1990; Manning 1990] rather than [MIH90, Man90] % % 1. Use semicolons between cites instead of commas % 2. Don't put hboxes around citation, so it can be broken across lines. \def\@citex[#1]#2{\if@filesw\immediate\write\@auxout{\string\citation{#2}}\fi \def\@citea{}\@cite{\@for\@citeb:=#2\do {\@citea\def\@citea{;\penalty\@m\ }\@ifundefined {b@\@citeb}{{\bf ?}\@warning {Citation `\@citeb' on page \thepage \space undefined}}% {\csname b@\@citeb\endcsname}}}{#1}} % Original Latex version: % \def\@citex[#1]#2{\if@filesw\immediate\write\@auxout{\string\citation{#2}}\fi % \def\@citea{}\@cite{\@for\@citeb:=#2\do % {\@citea\def\@citea{,\penalty\@m\ }\@ifundefined % {b@\@citeb}{{\bf ?}\@warning % {Citation `\@citeb' on page \thepage \space undefined}}% % \hbox{\csname b@\@citeb\endcsname}}}{#1}} % put cites in footnotesize unless in minipage (e.g. footnote) \def\@cite#1#2{{\if@minipage\else\footnotesize\fi[{#1\if@tempswa , #2\fi}]}} % DRAFT ANNOTATIONS: % Say \drafttrue near beginning to make draft remarks appear \newif\ifdraft \global\draftfalse % REMARKS are comments to draft readers; stars should appear in margins. \newcommand{\remark}[1] {\ifdraft{\it[\marginpar[{\hfill\Huge$\star$}]{{\Huge$\star$}}#1]}\fi} %% put marginpar after first char '[' so it will be vertically aligned. % ... draft header/footer annotations omitted \makeatother %% %% ---------------- Begin Document ---------------- %% \begin{document} % below controls draft notice on bottom of first page. \ifdraft\thispagestyle{draftplain}\draftfooter\else\thispagestyle{plain}\fi \title{Participatory Semantics Group} \author{\Large Progress Report 1995--1996} \maketitle %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% %% The following \staff macro and {people} environment were created %% especially for these progress reports. {people} is similar to %% {tabular}, except that it makes vertical columns instead of horizontal %% rows. (It uses TeX's \valign instead of \halign.) This is convenient %% because you can keep an alphabetical list of people, and easily break it %% into a new column in the appropriate place. (And CHANGE it easily too.) %% Put a & in front of each name---EXCEPT the first name in a column. %% Put a \\ after the last name in a column. %% Do not put a \\ after the last name in the entire list. %% See below. %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% \staff{Academic Staff} \begin{people} Carl Hewitt, Group Leader \end{people} \staff{Graduate Students} \begin{people} Miguel Hall &Carl Manning \end{people} \staff{Undergraduate Student} \begin{people} Francis Hsu \end{people} \staff{Support Staff} \begin{people} Marie Lamb \end{people} \newpage \begin{center}\LARGE\bf Know-How Infrastructures\\ for\\ Multi-Agency Systems \end{center} \markboth{}{} \begin{center} Carl Hewitt \& Carl Manning\\ hewitt@ai.mit.edu, carlmanning@ai.mit.edu\\ 16 May 1996 \end{center} %\centerline{\fbox{\psfig{file=Mobile-Distributed-logo.eps,height=5cm}}} \centerline{\psfig{file=Mobile-Distributed-logo.eps,height=4cm}} \section*{Abstract} \bgroup\small What is exciting about agents? One attraction is that agents promise to address one of the fundamental problems we currently have with our current information infrastructures: how to facilitate making semantic enhancements to improve assessing, predicting, planning, and robustness. However, the foundations of agent-oriented systems are in disarray. One significant problem is that in typical formulations, agents don't scale: an organization of agents is not an agent. The infrastructure for creating organizations is in its infancy. To address this issue, we introduce {\em agency} as a process which contributes toward rendering a service. Rendering a service is a discrete unit of activity that has value (that potentially can be sold). We present a conceptual methodology for understanding multi-agency systems, which focuses on interactions between participating agencies in space-time. We call this methodology Participatory Semantics. We are applying this methodology to multi-agency negotiations over future activity, where the participants must share and elicit potentially conflicting information. \egroup % small \ \newline Topic: Conceptual Foundations \section{Know-How Infrastructures} \boxfigure{ %Continually Available Info Infrastructure (not only at work stations) % enhances (robustness, manageability, accountability) by % enhances (availability, information, interactiveness, responsiveness) of % participatory semantics \caption{\bf Know-How Infrastructures: Synergy of Continual Availability \& Semantic Enhancement} \centerline{\psfig{file=khi-synergy.eps}}} {\em Know-How infrastructures} is our name for information infrastructures that are comprised of two interdependent and mutually supporting components: \begin{itemizeNoSpace} \item[1.] {\em Continually available telecomputing} provides information services at places and times where they are not available today. Information services are now mainly available only sporadically during the course of a day (e.g., while at a desk, at breaks in meetings). Know-How infrastructures will be large scale information infrastructures linking humans, equipment, and services using wired and wireless communication. \item[2.] {\em Semantic enhancement} is the integrated use of semantic methods to address issues of meaning in continually available telecomputer systems (e.g., simulating, reasoning, modeling methods). \end{itemizeNoSpace} Advances in continually available information infrastructure enhance the development of semantic enhancement and vice versa. Continual availability means telecomputing services can be present during more human activity, bringing those services ``closer to the action'' where semantic enhancements can be better informed, and can provide better services to the human participants in multi-agency activity. In particular, continual availability improves: \begin{itemizeNoSpace} \item {\em Accessibility:} at hand as needed \item {\em Immediacy:} recorded on site as it happens \item {\em Responsiveness:} effective in real time \item {\em Transparency:} integrated naturally into the course of work \end{itemizeNoSpace} \noindent Semantic enhancements provide methods that can improve the usefulness of continually available telecomputing for participating in and managing multi-agency activity. These semantic methods may use modeling, reasoning, and simulating to improve: \begin{itemizeNoSpace} \item {\em Assessing:} understanding past participation. \item {\em Predicting:} anticipating potential contingencies \item {\em Planning:} arranging future participating \item {\em Robustness:} dealing with conflict and difficulty \end{itemizeNoSpace} \noindent We call our approach to semantic enhancements {\em Participatory Semantics}. \boxfigure{ \caption{{\bf Emergency Management Vignette} ({\sl opportunities for the use of Know-How infrastructure emphasized})} \label{formal-methods-in-emergency-response} \bigskip\bigskip\small Consider an emergency management situation, such as battling forest fires. Organizations from several different jurisdictions and areas of expertise work together to understand and limit the fire's progress. In the meetings between these organizations, meteorologists may constructively analyze and forecast the weather ({\sl weather prediction methods}) to forecast weather e.g., winds and precipitation. Forest rangers may constructively analyze the path and speed of fires ({\em fire prediction methods}) in part by modeling the conditions and simulating the effects of the weather. {\sl Sensitivity analysis methods} are used to analyze plans for sensitivity to different likely contingencies --- shifts in the wind, faster progress through dry terrain, equipment breakdowns. Know-How infrastructures enable continual interaction among the teams and instruments operating in the field and air, and among live operations, constructive analysis, and visualization activities. Observations and readings update the maps and models; visualized maps of conditions, visualized models of the fire's progress, and visualized simulations of proposed plans, all inform the participants. As the weather and fire fighting teams encounter contingencies in the field, {\sl plan revision methods} are continually used to assess ongoing activities, reorganize plans, and relay them to other teams. Today continually available communications between teams is primarily voice through radio or phone; with Know-How infrastructure multimedia information can be sent digitally and processed locally. For example each team member can manipulate their own map view of deployment and conditions in neighboring areas; messages of a neighboring teammate in trouble can be prioritized over those of another team in a different valley. } \boxfigure{ \caption{{\bf Learning/Educating/Training Vignette} ({\sl opportunities for Know-How infrastructure emphasized})} \label{formal-methods-in-education-and-training} \bigskip\bigskip\small Know-How infrastructure can make project interactions explicit so participants can reflect on them and learn more from the experience. Learners can use continual availability of their Know-How infrastructure in labs, classrooms, tutorials, and at home. Consider a learning/educating/training situation, where students in geographically separated schools participate in joint projects such as managing school compost heaps and recycling programs and projecting what their likely future may be. Learners can measure the pH and temperature of school compost heaps and use these measurements to help understand how they operate. Analyzing these measurements illustrates important principles of biology, chemistry, and physics. Learners can model the circumstances of their own communities by semantic methods such as simulating/modeling \footnote{e.g., using systems that are generalizations of simple simulation interfaces such as SimCity, SimEarth, etc. in that all of the underlying models are transparent, which learners can modify and extend}. They compare and contrast with models and simulations developed by learners in other communities. In time they pass beyond this stage to create and program their own {\sl semantic} methods, and {\sl use their own data collection and analysis methods}. These semantic methods can be shared with learners in other communities. Projections of the future are uncertain, so learners experience how to probe for sensitivity to different contingencies --- shifts in population growth, industrial and agricultural effects, and changes in industrial production. A crucial aspect of projects like these is that they help {\it integrate} experience. Learners experience how to apply not only what they learn in mathematics, biology, chemistry, and physics but also what they learn in subjects such as politics and history. Mathematics, biology, chemistry, and physics are needed to understand how compost heaps work. Politics, economics, and history are needed to understand how compost heaps and recycling affect people's lives. Their own experiences interacting with learners of another community may itself provide background for their political and historical understanding. } Know-How infrastructures provide means for improving influence on future activities. For example, they can be used in activities such as assessing alternatives, training personnel, and monitoring competitiveness. Schedules, plans, action items, agendas, budgets, etc. can be used to propose alternatives, find conflicts, evaluate options, make agreements, propagate changes, etc. Consider the multi-agency vignettes in figures~\ref{formal-methods-in-emergency-response} and \ref{formal-methods-in-education-and-training}. Each vignette illustrates several opportunities for the use of Know-How infrastructure. Know-How infrastructure can be valuable during emergency response and recovery (figure~\ref{formal-methods-in-emergency-response}). In medicine, Peter Szolovits {\it et al.} are proposing the development of lifetime patient-centered health telecomputing systems {\cite{Szolovits1994}}. A key aspect is continual availability of medical telecomputing services, initially to outpatients such as diabetics. Participating effectively in project teams and partnerships is an important part of people's activities as lifelong learners (figure~\ref{formal-methods-in-education-and-training}). Teaming, partnering, and networking are important social aspects that can be loosely characterized as follows: \begin{itemizeNoSpace} \item In teaming, {\em projects} are given and team members work the projects. \item In partnering, {\em relationships} are given and partners develop the relationships. \item In networking, {\em encounters} are given and participants elaborate the encounters. \end{itemizeNoSpace} \noindent Activities will typically involve all of the above aspects. Know-How infrastructures technologies can make semantic methods explicit and available in meetings, in labs, and other times and places where telecomputing technologies are not now available. Know-How infrastructures are especially enabling for learning activities as learners spend their time on the move between classes, libraries, meeting rooms, laboratories, etc. \section{Participatory Semantics} \begin{quote} {\it ``Our intention with this initiative is to promote the notion that part of the reason for the repeated failure of large computing systems is that we (this culture) have reached the end of a paradigm in computing, and that we need more and different mathematics, indeed a new theoretical foundation for the science of computing, in order to proceed. The focus of this initiative is the nature of that new mathematics that might be important in a new foundation for computing.''} {\cite{Landauer&Bellman1994}} \end{quote} {\em Participatory semantics} is our approach to semantic enhancements for participation in Know-How infrastructures. Participatory semantics aims to provide conceptual and mathematical foundations to address how the meanings of participation can be described and processed using telecomputer services. The semantics of participation is very broad in that it includes the influences on all subsequent participation. In particular it includes social, psychological, cognitive, and affective aspects of meanings in participation. %% need to balance by citing other influences % %{\cite{Montalvo1995,Picard1995}} Semantic methods address issues of precision and accuracy in meaning. Representation technologies developed in the field of Artificial Intelligence (cf.~{\cite{DavisShrobe&Szolovits1993}}) are important precursors to semantic methods technologies. Work on semantic methods builds on previous work on open information systems summarized in {\cite{Hewitt&Inman1991}}. For example {\cite{Manning1987}} describes a debugging system constructed on precursor principles of semantic methods. One important characterization of semantic methods is terms of preparation, performance, and assessment: \begin{itemizeNoSpace} \item {\sl Preparation methods} for preparing for future participation. Preparation is multi-faceted in that it can contribute to many different future activities. \item {\sl Performance methods} for participation. Performance includes preparation and assessment as special cases. \item {\sl Assessment methods} for assessing previous participation. Assessment is multi-faceted in that it can assess many different previous activities. \end{itemizeNoSpace} \remark{it would be nice to follow up on this --- if this is important, can we do it on our LAPD or other example later?} \subsection{Accountings, Space-Time Regions, Attributions} The participatory semantics approach to semantic methods begins by formulating how we can develop accountings of participation in multi-agency activity. Criteria include: \begin{itemize} \item Concurrency: deals with distributed, open, multi-agency systems \item Scalability: provides modularity and abstraction to manage increasing size and complexity. \item Plurality: deals with multiple accountings of what was, is, or will be happening, which may conflict, or may be incommensurable, but are always incomplete. \end{itemize} We present 4 dimensional space-time accountings of activity in 2 dimensional diagrams such as figure~{\ref{LAPD-fig}}. These presentations are composed of: \begin{itemize} \item Accountings of activities (square-cornered rectangles containing diagrams) \item Space-time regions \begin{itemizeNoSpace} \item Activities (round-cornered rectangles) \item Participants (horizontal "lifelines") \item Communication paths (solid-line arrows) \end{itemizeNoSpace} \item Attributions (text with named arguments, or dotted-line arrows) \end{itemize} An {\em accounting} tells the story of some activity. Like a good news story, the accounting may contain multiple descriptions (views) of the same activity: A summary overview, description placing activity within surrounding context, and detailed accounts of the activity. \begin{figure}[p] \caption{An Accounting. (page 1 of 2)} %% Note: for some reason the BoundingBox in eps files was not accurate. %% edited x dim so changed BoundingBox from %%BoundingBox: 0 0 658 756 %% to %%BoundingBox: 54 0 604 756 %% in order to get figure to fill desired width. \label{LAPD-fig} \psfig{file=LAPDA-P1.EPS,width=\textwidth} \end{figure} \begin{figure}[p] \addtocounter{figure}{-1} \caption{(continued) (page 2 of 2)} \psfig{file=LAPDA-P2.EPS,width=\textwidth} \end{figure} \begin{quote} For example, the RescuedInLA.account contains \begin{itemizeNoSpace} \item a simple summary (RescuedInLA) and \item a more detailed account (RescuedInLA.detailed) \item and it places the activity in context (RescuedInLA.context) \end{itemizeNoSpace} There are also more detailed views of subregions (titled Interaction\#1.detail, Interaction\#2.detail, etc.) \end{quote} \noindent An accounting divides space-time into {\em space-time regions} where activities take place. The accounting may divide a region into subregions and relate those subregions. \begin{quote} For example, \begin{itemizeNoSpace} \item in RescuedInLA.context, being StrandedInLA is separated in space from the LAPD. \item Within RescuedInLA.detail, Interaction\#1 is separated in time from Interaction\#2 by a region in which the strandee waited. \end{itemizeNoSpace} \end{quote} \noindent {\em Participants} have space-time regions during which they exist, live, and participate in activities, their {\em lifelines}. Participants' lifeline regions intersect with the space-time regions of the activities in which they participate. By convention, time in a diagram increases from left to right, so later activities are diagrammed to the right of earlier activities on a lifeline. \begin{quote} For example, in RescuedInLA.context, the LAPD participated in several activities including Interaction\#1, then later Interaction\#2, and later recorded the activities. \end{quote} \noindent {\em Communications} exist in and travel through space-time. The path of a message is a region of space-time. \begin{quote} For example, the communications in Interaction\#1.detail connect the Strandee and the LAPD in space-time. They are ordered by the sending and receiving activities on the two lifelines. \end{quote} \noindent An accounting may itself contain nested accounts of past events, future plans, or both. \begin{quote} For example, Message\#1 and Message\#2 contain plans; these plans involve historical context of the situation and future activities. \end{quote} \noindent An {\em attribution} expresses a relationship. When an Attribution is shown in a space-time region, it means that someone has attributed that the relationship specified by the attribution holds in that region. The region is an implicit argument in the attributed relationship. \begin{quote} StrandedInLA is attributed to having been a region where and when the Strandee waited. It is also a region attributed to being a region where the Strandee was stranded by the elevator in Marina del Rey. \end{quote} \noindent Whether an attribution holds throughout the region or says something exists within the region depends on the attribution. \begin{quote} During the region called StrandedInLA, the Strandee was stranded the entire time. However, in the region Interaction\#1, the postponement didn't happen until the end of the conversation; it did not occupy the entire region. \end{quote} In these space-time diagrams, only topological information is significant, i.e., the important information is in which regions are nested in, connected to, ordered by, etc. other regions and attributions. The relative sizing, spacing, and alignment (except for sharing a border) is not significant. \begin{quote} For example, in RescuedInLA.detail, the relative widths of Interaction\#1 and Interaction\#2 is not any indication of their relative durations. Nor is it important that the Strandee and LAPD enter RescuedInLA.detail at the same horizontal position (this does not mean they entered the activity at the same time). It is significant that the wait between Interaction\#1 and Interaction\#2 fills the time between them so their borders meet; it is not significant that the 1hr wait takes less width on the page than either of the interactions. \end{quote} \noindent The absence of information in a diagram does not imply it is not true. For example, if two regions do not overlap in the diagram, this does not necessarily mean that they are known not to overlap.\footnote{However, putting in all this information in the diagram would obfuscate the diagram. Dealing with this presentation issue is an area of future work.} % One approach is to use defaults --- if % we create a diagram in which two regions do not overlap, then by % default we may assume they do not overlap, but be prepared to % reconsider that assumption. Similarly, if we don't know whether two % regions overlap, we may draw them without overlapping by default, % but be prepared to revise the diagram.} \section{Multi-Agency Systems (MAS)} {\em Agencies} is the name that we use for the means by which services are carried out in telecomputer systems. {\em An Agency is a process which contributes toward rendering a service.} An agency viewpoint concentrates on {\em how} systems work as opposed to focusing on exactly which individuals participate. {\em Rendering a service is a discrete unit of activity that has value (that potentially can be sold).} Making extensive use of telecomputing services inevitably means that multiple agencies interact with each other. In figure~\ref{LAPD-fig}, the LAPD is an agency, an ongoing process which provides a service of responding to emergency 911 calls; exactly who is answering which call is not important. The LAPD is internally complex to manage the services it provides with the resources it has. As in the example when the strandee wasn't able to get help for several hours because the LAPD was attending to the riots, the LAPD may not always have resources dedicated to each service, and response may be delayed as a result. We can also view the guard and elevator system as agencies which contributed to freeing the Strandee: the guard rendered the service of activating the elevator, and the elevator system (once activated) rendered the service of transporting the Strandee to the first floor. The communications system also renders services, enabling the Strandee to communicate with the LAPD (but not, in this case, to communicate directly with the guard). Consider the agencies involved in helping route calls to emergency services during disaster response and recovery. In this case different emergency agencies (fire companies, hospitals, EMTs, police, etc.) can make use of semantic methods (e.g., to communicate location of fire trucks, availability of beds, availability of medical personnel, etc.). When organizations are overloaded with calls, receptionist agencies may consult online representatives of nearby organizations to find alternatives and who has the most appropriate resources available. For example, when a call came in during the L.A.~riot regarding a trapped victim who was not in danger, police resources were not immediately available. While police resources were completely occupied, other organizations such as nearby building security guards had resources available in the building. Semantic enhancements enable participants to make meaningful original contributions (such as constructing and suggesting alternative plans), or to process contributions with less burden on human participants (e.g., prioritizing requests such as the original help form, fulfilling information processing requests such as finding how to contact the guard on duty and forwarding the plan). % If % receptionist services had forwarded the call to neighborhood negotiation % spaces, participants in these spaces could recognize that security % guards were available, and the security guards could have been % contacted much more quickly {\cite{Hewitt&Manning1994}}. There are several advantages to analyzing large scale interactions in terms of agencies rather than agents: \begin{itemize} \item Agency allows the internal processes (agencies and individuals) to themselves be implemented as distributed, concurrent agencies. Typical agent formulations and architectures limit the kinds of internal concurrency. For example, SOAR {\cite{Newell1990}}) is limited to one goal stack. \item Conversely, combining and organizing agencies results in another (composite) agency. In typical agent formulations and architectures, the combination of two or more agents is not an agent. E.g., in the SOAR architecture, the combination of two SOAR agents is not itself an agent and Society of Mind architecture is not itself a Society of Mind. (there are no K-Lines between them). \end{itemize} \remark{need more here!} % Agency abstracts the behavior and responsibilities of % systems of subagencies and individuals. While an agent may % serve as a representative (or interface) for an agency, % abstracting the behavior of the agency as the behavior of % the representative limits the behavior (e.g., the representative % may become a bottleneck) \boxfigure{ \caption{\bf Multi-Agency Systems} \centerline{\psfig{file=3way-wireless.eps,height=5cm}}} \floater{The road to ``agent'' hell will be paved with {\em lots of helpful intentions.}} \section{Negotiating in Know-How Infrastructure} Participants in Know-How infrastructure have limited knowledge of each other's activities. To arrange and perform activities, participants can share information. Where their activities conflict, participants can explore alternatives. Also they can explore to improve the outcomes for everyone. This process where multiple participants interact to influence future participation, can be characterized as negotiation. \floater{Negotiation is largely about influencing future participation.} \begin{quote} For example, in figure~\ref{LAPD-fig} the strandee made a request for help; the LAPD elicited further details (danger, problem, location), and found that immediate action conflicted with its current first priority, dealing with the riots, so it constructed its best quick alternative (waiting 4hrs) and left the negotiation. The strandee explored other alternatives, and reopened the negotiation with a workable plan which improved the outcome for both of them: the strandee was freed earlier, and the LAPD had one less case to deal with. \end{quote} An important opportunity for the use of participatory semantics is in negotiating the meanings of telecomputer participation in Know-How infrastructures. Semantic enhancements may be used to communicate about interdependent activities, to explore the influences of opportunities, to reveal possible contingencies and ways to avoid conflicts, and to create and adjust agreements and plans as a result. \begin{quote} For example, in figure~\ref{LAPD-fig} the LAPD sent a plan for a 4 hour delay with enough context to explain the situation. The strandee may have used local services to further analyze the situation and construct a possible but incomplete alternative, hoping the LAPD agency can contact the guard from the building description given, and hoping the guard is able to have the elevator activated. \end{quote} \noindent Semantic methods may be used to interact with an incoming communication, so that telecomputing services may analyze the opportunities and constraints (e.g., deadlines, conflicts), reveal contingencies (e.g., prioritize against other tasks), and adjust plans (e.g., schedule an interruption to deal with the communication; see~\S\ref{sec:interruption-management}). Participants can use negotiation semantics to influence many aspects of future participation including: \begin{itemizeNoSpace} \item {\em tasks} to be performed (e.g., breakdown of project, assignments of action items), \item {\em products} to be consumed, created, or compared (via verbal and textual descriptions, pictures and diagrams, photographs, videos, etc.). \item {\em resources} to be used (including available human and equipment time, as well as stocks to be used), and \item {\em dependencies} between them (including budget and schedule restrictions, as well as dependencies between inputs and outputs). \end{itemizeNoSpace} Participatory semantics is concerned not only with participants, tasks, and resources, but also the dependencies and other {\em relationships} between them. These relationships include those between parts of different departments or organizations, so the most up-to-date information may be {\em distributed} on several individuals' and organizations' telecomputing systems. \begin{quote} For example, the strandee did not know about the riot or the LAPD's task of dealing with it; only when the strandee contacted the LAPD, and the LAPD contributed the delay plan\#1 in context, did the strandee learn that a non-urgent request to be rescued conflicted with, and had lower priority than, the LAPD's ongoing task of dealing with LA riots. (More traditional plan dependencies, such as needing to activate the elevator before the strandee can ride it down, could also be added as attributions between these activities.) \end{quote} Semantic methods give telecomputing services more information about participation, enabling them to provide additional services (cf.~{\cite{Dertouzos1991}}). Thus, people may have semantic methods for their own tasks on their own telecomputing systems, linked through networks with semantic methods for managing project plans on the telecomputing systems of their organization and perhaps those of other organizations. \begin{quote} For example, if the strandee was using a telecomputer to keep an appointment calendar and telephone log, the telecomputer would already have the location and call information shown in the accountings. \end{quote} \subsection{Negotiation Systems for Participatory Semantics} To deal with the requirements of Know-How infrastructures, we are developing {\em negotiation systems} to enable multiple agencies to ``plug-in'' to negotiation services. Like a hardware bus, a negotiation system provides a protocol for software services (like hardware boards) for negotiation about the meanings of participation, cf.~the software bus in {\cite{BellmanGillam&Landauer1993}}. Unlike early hardware buses (such as the ISA bus of the PC world), negotiation systems should allow other software services to be added easily without changing other participants e.g., dip switch settings. Like more recent busses such as the PCMCIA bus, it should automatically perform any reconfiguration needed to enable participants to be plugged in or removed while the system is in operation. The wrapping approach of Landauer \& Bellman {\cite{Landauer&Bellman1993}} can be used to integrate other software systems using negotiation systems. A hardware bus enables cooperation by requiring each board to make requests and receive permission before taking any bus action. Negotiation systems can also require participant systems to request permission by making proposals and requesting the approval of other systems before taking action under the authority of a negotiation system. However, hardware bus protocols manage a simple resource (access to the bus to transfer a short data or signal burst in the immediate future) and thus can get by with protocols composed of a few request, priority, grant, acknowledge, etc. signals. A negotiation system may manage resources and resource use patterns with much greater complexity by providing participants with semantic methods for managing resources, their relationships, constraints, behavior, and interactions. Semantic negotiation spaces can make use of the following semantic methods: \boxfigure{ \caption{\bf Negotiation Systems} \centerline{\psfig{file=3way-forum.eps,height=5cm}}} {\em A negotiation system can increase robustness by referring problematic cases to other parties.} If the negotiation peters out or times out without a clear agreement, then the negotiation so far may be summarized and passed on to other parties (e.g., a human, or another negotiation system) for more in-depth processing. Thus, some unresolved conflicts are detected and brought to the attention of other parties. {\em Negotiation systems can provide status reports.} At their own initiative negotiation systems can provide status reports to other parties. In addition other parties can request status reports. \subsection{Overload and Interruption management} \label{sec:interruption-management} Know-How infrastructures can be continually available through mobile devices and wireless networks. Continually availability raises important issues, such as the tradeoff between availability and the continual threat of being interrupted by incoming calls or messages. Effectively managing notifications is an important requirement for Know-How infrastructures. Notifications are any asynchronous output to a human (including aural, visual, tactile, etc.). Notifications may be generated not only as a result of incoming calls and messages, but also as a result of ongoing processes within the Know-How infrastructures. \floater{Continual Availability raises important issues, such as the tradeoff between availability and the continual threat of being interrupted by incoming calls or messages.} \boxfigure{ \caption{\bf Interruption and Overload Management} \centerline{\psfig{file=interrupt-mtg.eps,height=5cm}}} Know-How infrastructures can enable people to communicate with each other and access remote services to improve coordination of their participation. However, even today many people are inundated with \begin{itemizeNoSpace} \item overflowing email inboxes from mailing lists in which they are only partly interested, obscuring important messages; \item phone calls and pagers causing inopportune interruptions in meetings and other activities, just because calls {\em could} be important \end{itemizeNoSpace} \noindent With continual availability, communications can arrive at any time, potentially disrupting ongoing activity, whether it is important/urgent/relevant or not. As people start using communications during their meetings (e.g., to exchange proposals and documents, to consult partners and team members away from the meeting, to invoke remote telecomputing services, etc.), it is less of an option to just shut down communications during the meeting. Many organizations hire staff to deal with some of these problems, but while these are becoming issues for everyone, not everyone can afford to hire human staff to deal with them. Thus, for example, staff may filter calls so as to only interrupt a meeting for matters more important and urgent than the meeting. Human staff take into account the status of current activities of their employer, the importance of their concerns about different aspects of those activities, and the relationships of other actors to the employer and those activities. Already some people make use of rule systems for sorting and prioritizing email based on header fields and keywords. However, people change these rules slowly; when people use Know-How infrastructures to help plan and track their activities and relationships, the opportunity and need arises to use ongoing activity information as well to filter and prioritize the communications received, relative to the activity at hand (e.g., an ongoing meeting). One approach is to build on the semantic methods people already use to manage their participation (e.g., in electronic appointment books, project management tools, etc.). Knowledge of the the owner's plans and habits can be used to estimate the owner's likely current activity. (Sensing location and who else is in the same room can provide more clues to the owner's current activity.) Semantic methods provide a basis for personal receptionist services. The service tracks different aspects of a person's activities (e.g., projects, customers, career, family, etc.). When an incoming phone call arrives, the receptionist service may consult the owner's schedule to help find out if the owner is available for the call. When the owner is busy (e.g., in a project team meeting), additional information about the call may be elicited through negotiation with the caller, or with the caller's personal services. Simple but helpful call information includes the sender's identity, the intended recipient (personal or organization/list), and/or the subject (e.g., a sales call, a subordinate calling in sick). This information about the call may help the services associate the call with recipient's activities (e.g., it's a call about an important part the team needs soon for another project). The service may consider alternative ways of handling the call (take a message, forward the call to an associate, interrupt the meeting), and reasons for prefering one alternative to the next as a way of narrowing down which response is appropriate in the context of the current activities. % \section{Future Work} \section{Conclusion} \boxfigure{ % \caption{} \centerline{\psfig{file=ss-paradigm.eps,width=\boxwidth}}} {\em Know-How infrastructures} enable people and agencies to invoke participatory semantics methods without interrupting ongoing activities. One promise of agency technology is the semantic enhancements they provide to information infrastructure. Semantic enhancements can improve operations by reducing time, cost, and risk in designing, testing, planning, analysis, coordination, and assessment activities among multiple agencies. In the participatory semantics approach to semantic enhancement, we account for {\em agency} as a process which contributes toward rendering a service. Rendering a service is a discrete unit of activity that has value. We have developed a methodology for accounting for agency activity in space-time, which deals with \begin{itemizeNoSpace} \item concurrency: participatory semantics accountings divide space-time into regions, and attributions are made within those regions (cf. not about global states or worlds) \item scalability: agencies working together may be modularized as a larger, more complex, agency, and attributions may be made abstracting the services rendered in terms of the larger agency. \item plurality: an accounting may provide multiple views of a agency, and may itself contain multiple accountings by different authors. \end{itemizeNoSpace} Important issues for semantic enhancements for Know-How infrastructure include the following: \begin{itemizeNoSpace} \item {\em semantic methods} for addressing issues of precision and accuracy in processing the meanings of participation in Know-How infrastructures, \item {\em services and standards} for negotiation spaces, \item {\em socially acceptable paradigms and technologies} for continual availability, \item {\em interfaces} to assist interaction with Know-How infrastructures and their semantic methods. \end{itemizeNoSpace} \section{Acknowledgments} The work for this paper was sponsored in part by ARPA through ONR and IDA and in part by MIT. The views expressed in this paper are not necessarily those of ARPA, ONR, IDA, the US government, or MIT. Those who know Kirstie Bellman, Randy Garrett, Chris Landauer, and Julia Loughran will find their voices in much of what appears in this paper. The conceptual development of semantic methods has been done jointly with Kirstie and Chris. The conceptual development of Know-How infrastructures has been collaborative work with Randy, Julia, and the other participants in the ADS TAB. We look forward to further participation with all of them in future work on semantic methods for Know-How infrastructures. Discussions with David Brock, Peter Brooks, Gary Coe, Dale Lichtblau, Tom Knight, H.T. Kung, Fanya Montalvo, and Mike Wessler were very helpful in developing this paper. \begin{secbibliography}{999} \bibitem[Agha 1986]{agha-book} Gul Agha, {\em Actors: a Model of Concurrent Computation in Distributed Systems}. Cambridge, MA: MIT Press, 1986. \bibitem[Asai, Matsuoka, \& Yonezawa]{AsaiMatsuoka&Yonezawa1996} ``Duplication and Partial Evaluation-For a Better Understanding of Reflective Languages'', in {\em Lisp and Symbolic Computation}, (special issue on Computational Reflection Guest Editors: Daniel P. Friedman and Akninori Yonezawa), pp.~203--242, vol.~9, no.~2/3, May/June 1996. \bibitem[Barbuceanu \& Fox 1995]{Barbuceanu&Fox1995} Mihai Barbuceanu and Mark S. Fox, ``COOL: A Language for Describing Coordination in Multi Agent Systems'', in {\em ICMAS-95 Proceedings of the First International Conference on Multi-Agent Systems}, San Francisco. Cambridge, USA: AAAI/MIT Press 1995, p17--24 \bibitem[Bellman 1993]{Bellman1993} Kirstie L. Bellman, ``The Challenges in Testing Heterogeneous Software Environments'', Proceedings AIAA Computers in Aerospace 9 Conference, 19--21 October 1993, San Diego 1993. \bibitem[Bellman, Gillam, \& Landauer 1993]{BellmanGillam&Landauer1993} Kirstie Bellman, April Gillam, and Christopher Landauer, ``Challenges for Conceptual Design Environments: The VEHICLES Experience'', {\em Revue Int. de CFAO et Infographic}, (Paris), V.8(3), Fall 1993. \bibitem[Carver \& Lesser 1995]{Carver&Lesser1995} Norman Carver and Victor Lesser, ``The DRESUN Testbed for Research in FA/C Distributed Situation Assessment: Extensions to the Model of External Evidence'', in {\em ICMAS-95 Proceedings of the First International Conference on Multi-Agent Systems}, San Francisco. Cambridge, USA: AAAI/MIT Press 1995, p33 \bibitem[Castelfranchi 1995]{Castelfranchi1995} Cristiano Castelfranchi, ``Commitments: From Individual Intentions to Groups and Organizations'', in {\em ICMAS-95 Proceedings of the First International Conference on Multi-Agent Systems}, San Francisco. Cambridge, USA: AAAI/MIT Press 1995, p41--48 \bibitem[Cimatti \& Serafini 1995]{Cimatti&Serafini1995} Alessandro Cimatti and Luciano Serafini, ``Multiagent Reasoning with Belief Contexts II: Elaboration Tolerance'', in {\em ICMAS-95 Proceedings of the First International Conference on Multi-Agent Systems}, San Francisco. Cambridge, USA: AAAI/MIT Press 1995, p57--64 \bibitem[Clinger 1981]{Clinger1981} Will Clinger, {\em Foundations of Actor Semantics}, University Microfilms (Ann Arbor, Michigan) MIT Doctoral Dissertation in Mathematics, June 1981. \bibitem[Cohen \& Levesque 1995]{Cohen&Levesque1995} Philip R. Cohen and Hector J. Levesque, ``Communicative Actions for Artificial Agents'', in {\em ICMAS-95 Proceedings of the First International Conference on Multi-Agent Systems}, San Francisco. Cambridge, USA: AAAI/MIT Press 1995, p65--72 \bibitem[Conte \& Castelfranchi]{Conte&Castelfranci1995} Rosaria Conte and Cristiano Castelfranchi, {\em Cognitive and Social Action}, UCL Press (London, England) 1995. % \bibitem[Berners-Lee 1993]{Berners-Lee1993} % Tim Berners-Lee, ``Uniform Resource Locators (URL): A Unifying Syntax for % the Expression of Names and addresses of Objects on the Network'', % IETF Working Draft, October 1993. % \bibitem[Bhagwat \& Perkins 1993]{Bhagwat&Perkins1993} % Pravin Bhagwat and Charles E. Perkins, % ``A Mobile Networking System based on Internet Protocol (IP)'', % USENIX Symposium on Mobile and Location Independent Computing, % p.69--82. Cambridge, MA, August 1993. % \bibitem[Blackwell et al 1994]{Blackwell1994} % T. Blackwell, K. Chan, K. Chang, T. Charuhas, J. Gwertzman, B. Karp, % H. T. Kung, W. D. Li, D. Lin, R. Morris, R. Polansky, D. Tang, C. % Young, and J. Zao, ``Secure Short-Cut Routing for Mobile IP,'' Conference % Proceedings of Usenix Summer 1994 Technical Conference, Boston, % Massachusetts, June 6--10, 1994, pp. 305--316. % \bibitem[Cattell 1994]{Cattell1994} % R.G.G. Cattell, {\em Object Data Management (revised edition)}. % Reading, MA: Addison-Wesley, 1994. \bibitem[Davis, Shrobe, \& Szolovits 1993]{DavisShrobe&Szolovits1993} Randall Davis, Howard Shrobe, and Peter Szolovits, ``What Is a Knowledge Representation'', {\em AI Magazine}, 14(1):17--33, Spring 1993. \bibitem[Decker 1995]{Decker 1995} Keith Decker, ``Distributed Artificial Intelligence Testbeds'', Tutorial Notes, ICMAS-95. \bibitem[Dertouzos 1991]{Dertouzos1991} Michael L. Dertouzos, ``Communications, Computers, and Networks'', {\em Scientific American}, (special issue), vol.~265, no.~3, p.~62--69, September~1991. \bibitem[Ferber 1995]{Ferber1995} Jacques Ferber, ``Multi-Agent Tools: Platforms, Actors, Reflection'', Tutorial Notes, ICMAS-95. \bibitem[Gasser 1995]{Gasser1995} Les Gasser, ``Computational Organization Research'', in {\em ICMAS-95 Proceedings of the First International Conference on Multi-Agent Systems}, San Francisco. Cambridge, USA: AAAI/MIT Press 1995, p414-415. \bibitem[Gasser, Rosenschein, \& Ephrati 1995]{GasserRosenschein&Ephrati1995} Les Gasser, Jeffrey S. Rosenschein, and Eithan Ephrati, ``Introduction to Multi-Agent Systems (and Distributed Artificial Intelligence)'', Tutorial Notes, ICMAS-95. \bibitem[Hewitt 1985]{Hewitt1985} Carl Hewitt, ``The Challenge of Open Systems,'' {\em Byte}, Vol.~10, No.~4, April 1985, pp.~223-242. \bibitem[Hewitt \& Inman 1991]{Hewitt&Inman1991} Carl Hewitt and Jeff Inman, ``DAI Betwixt and Between: From ``Intelligent Agents'' to Open Systems Science.'', {\em IEEE Transactions on Systems, Man, and Cybernetics}, Vol.~21, No.~6, November 1991. \bibitem[Hewitt \& Manning 1994]{Hewitt&Manning1994} Carl Hewitt and Carl Manning, ``Negotiation Architecture for Large-Scale Crisis Management Using NII'', AAAI-94 Workshop on Models of Conflict Management in Cooperative Problem Solving. Seattle, WA, August 4, 1994. p.53--68. \bibitem[Honda \& Tokoro]{HondaTokoro1996} Yasuaki Honda and Mario Tokoro, ``Reflection and Time-Dependent Computing: Experiences with the $R^{2}$, in {\em Lisp and Symbolic Computation}, (special issue on Computational Reflection Guest Editors: Daniel P. Friedman and Akninori Yonezawa), pp.~243--277, vol.~9, no.~2/3, May/June 1996. \bibitem[Huntback, Jennings, \& Ringwood 1995]{HuntbackJennings&Ringwood1995} Matthew M. Huntbach, Nick R. Jennings, and Graem A. Ringwood, ``How Agents Do It in Stream Logic Programming'', {\em ICMAS-95 Proceedings of the First International Conference on Multi-Agent Systems}, San Francisco. Cambridge, USA: AAAI/MIT Press 1995, p177--184 \bibitem[Ishida 1995]{Ishida1995} Toru Ishida, ``Parallel, Distributed and Multi-Agent Production Systems---A Research Foundation for Distributed Artificial Intelligence'', in {\em ICMAS-95 Proceedings of the First International Conference on Multi-Agent Systems}, San Francisco. Cambridge, USA: AAAI/MIT Press 1995, p416--422. \bibitem[Jefferson \& Friedman]{Jefferson&Friedman1996} Stanley Jefferson and Daniel P. Friedman, ``A Simple Reflective Interpreter'', in {\em Lisp and Symbolic Computation}, (special issue on Computational Reflection Guest Editors: Daniel P. Friedman and Akninori Yonezawa), pp.~181--202, vol.~9, no.~2/3, May/June 1996. % \bibitem[Katz 1990]{Katz90} % Randy H. Katz, ``Toward a Unified Framework for Version Modeling in % Engineering Databases'', {\em ACM Computing Surveys}, V22n4, December 1990. % \bibitem[Kowalski 1983]{Kowalski83} % Kowalski, R. A. ``Logic Programming'', % In {\em 1983 IFIP Congress Proceedings}, IFIP, August 1983. \bibitem [Landauer \& Bellman 1992]{Landauer-Bellman92} Christopher Landauer and Kirstie L. Bellman, ``Integrated Simulation Environments'' (invited paper), {\em Proceedings of DARPA Variable-Resolution Modeling Conference}, 5-6 May 1992, Herndon, Virginia, Conference Proceedings CF-103-DARPA, published by RAND (March 1993) \bibitem[Landauer 1993]{Landauer1993} Christopher Landauer, ``Applications of the Wrapping Approach in Software Engineering'', Proceedings AIAA Computers in Aerospace 9 Conference, 19-21 October 1993, San Diego 1993. \bibitem[Landauer \& Bellman 1993]{Landauer&Bellman1993} Christopher Landauer and Kirstie Bellman, ``The Role of Self-Refer\-ential Logics in Software Architecture Using Wrappings'', Proceedings of 1993 Irvine Software Symposium (ISS'93) Irvine, CA, April 30, 1993. \bibitem[Landauer \& Bellman 1994]{Landauer&Bellman1994} Christopher Landauer and Kirstie Bellman, ``New Mathematical Foundations for Computer Science, Workbook Edition 1: `Post-IMACS, Web Home Page, On-line Conference, MUD'\,'', http://www.cs.umd.edu/\~~cal/newmath.html \bibitem[Lux \& Steiner 1995]{Lux&Steiner1995} Andreas Lux and Donald Steiner, ``Understanding Cooperation: An Agent's Perspective'', in {\em ICMAS-95 Proceedings of the First International Conference on Multi-Agent Systems}, San Francisco. Cambridge, USA: AAAI/MIT Press 1995, p261--268. \bibitem[Malenfant, Dony, \& Cointe]{MalenfantDony&Cointe1996} Jacques Malenfant, Christope Don, and Pierre Cointe, ``A Semantics of Introspection in a Reflective Prototype-Based Language'', in {\em Lisp and Symbolic Computation}, (special issue on Computational Reflection Guest Editors: Daniel P. Friedman and Akninori Yonezawa), pp.~153--180, vol.~9, no.~2/3, May/June 1996. \bibitem[Manning 1987]{Manning1987} Manning, C. ``Traveler: the Actor Observatory'', In {\em 1987 European Conference on Object Oriented Programming}, Springer Verlag Lecture Notes in Computer Science series, Vol.~276. \bibitem[Micali 1994]{Micali1994} Micali, Silvio. {\em CS Proofs}, MIT/LCS/TM-510, June 1994. \bibitem[Minsky 1985]{Minsky1995} Marvin Minsky, ``Society of Mind'', Simon \& Schuster, 1985. \bibitem[Montalvo 1995]{Montalvo1995} Montalvo, Fanya, ``Emotional Computing'', part of Jan. 1995 MIT IAP Lecture Series on ``Emotions in the Workplace''. http://tinylondon.ucsd.edu/\~~cal/Fanya\_Montalvo.html \bibitem[Negroponte 1991]{Negroponte1991} Nicholas P. Negroponte, ``Products and Services for Computer Networks'' {\em Scientific American}, (special issue), vol.~265, no.~3, p.~106--113, September~1991. \bibitem[Newell 1990]{Newell1990} Allen Newell, {\em Unified Architectures of Cognition}, Harvard University Press, 1990. %\bibitem[NSTC CIC 1995]{NSTC CIC 1995} % National Science and Technology Council, Committee on Information and % Communication. % ``Strategic Planning Document --- Information and Communications'' % March 10, 1995. Available online through {\tt http://www.hpcc.gov}; % hardcopy at National Coordination Office for HPCC, % building 38A, Room B1N30, 8600 Rockville Pike, Bethesda, MD 20894 \bibitem[Picard 1995]{Picard1995} R. W. Picard {\em Affective Computing} MIT Media Laboratory Perceptual Computing Section Technical Report No. 321. http://www.media.mit.edu/\~~picard/ \bibitem[Rao \& Georgeff 1995]{Rao&Georgeff1995} Anand S. Rao and Michael P. Georgeff, ``BDI Agents: From Theory to Practice'', {\em ICMAS-95 Proceedings of the First International Conference on Multi-Agent Systems}, San Francisco. Cambridge, USA: AAAI/MIT Press 1995, p312--319. \bibitem[Sycara 1995]{Sycara1995} Katia P. Sycara, ``Coordination of Multiple Intelligent Agents in the Infosphere'', {\em ICMAS-95 Proceedings of the First International Conference on Multi-Agent Systems}, San Francisco. Cambridge, USA: AAAI/MIT Press 1995, p432. \bibitem[Szolovits 1994]{Szolovits1994} Peter Szolovits, Jon Doyle, William J. Long, Isaac Kohane, and Stephen G. Pauker, {\em Guardian Angel: Patient-Centered Health Information Systems} MIT/LCS/TR-604, May 1994. \bibitem[Tesler 1991]{Tesler1991} Lawrence G. Tesler, ``Networked Computing in the 1990s'', {\em Scientific American}, (special issue), vol.~265, no.~3, p.~86--93, September~1991. % \bibitem[Weider \& Deutsch 1993]{Weider&Deutsch1993} % Chris Weider and Peter Deutsch, ``Uniform Resource Names'', % IETF Working Draft, October 1993. \bibitem[Weiser 1993]{Weiser1993} Mark Weiser, ``Some Computer Science Issues in Ubiquitous Computing'', {\em CACM}, Vol.~36, no.~7, July 1993 \bibitem[Weiser 1991]{Weiser1991} Mark Weiser, ``The computer for the 21st century'' {\em Scientific American}, (special issue), vol.~265, no.~3, p.~94--104, September~1991. \bibitem[Wooldridge \& Jennings 1995]{Wooldridge&Jennings1995} Michael Wooldridge and Nick Jennings, ``Agent Theories, Architectures, and Languages'', Tutorial Notes, ICMAS-95. \bibitem[Yokoo 1995]{Yokoo 1995} Makoto Yokoo, ``Search Foundations for MultiAgent Systems'', Tutorial Notes, ICMAS-95. \end{secbibliography} \ifdraft \newpage \tableofcontents \fi \end{document}