Turtles: Protocol-Based Foundations for Distributed Multiagent Systems

There is growing interest in distributed systems and architectures whose components are autonomous social parties such as humans and organisations. The parties in such systems interact with each other via their software agents for the purposes of exchanging information and services. The interactions normally take the form of conversations (as opposed to invocations) realised over asynchronous messaging. Naturally, a crucial area of study for computer science and software engineering is the specification and enactment of interaction protocols, that is, the rules of encounter by which parties in the system would interact. Considered as such, the notion of protocol represents a generalisation of the notion of "contract" advocated in Design by Contract approaches.

A key question concerns the nature of contracts. Work in areas such as concurrency and Web services, has predominantly conceptualised protocols in terms of message ordering and occurrence constraints that must be respected by the parties' agents. We refer to such protocols as messaging protocols. Although messaging protocols serve the important purpose of distributed coordination, considered as contracts, they are too low-level for multiagent settings of autonomous parties. Specifically, they do not capture social constraints such as the commitments that are binding on the parties in the interaction.

This gap represents a substantial opportunity. In real life, commitments in fact represent the atoms of what people normally understand as contracts. Commitments accommodate the balance between, on the one hand, autonomy and flexibility, and, on the other hand, correct behaviour. Commitment specifications capture stakeholder requirements in multiparty domains. Further, the states of commitments underlie most key performance indicators (KPIs) that stakeholders are interested in any multiagent domain. Work in commitment protocols in multiagent systems has made progress in developing computational abstractions for commitments. However, important challenges related to expressiveness and distributed enactment of commitment protocols have not even been adequately formulated, let alone tackled. Ensuring correct distributed enactment for expressive commitment protocols is crucial to realising the full value of commitments as a human-level architectural abstraction.

The broad objective of Turtles is to bring commitment-based contracts to distributed computing. This project develops foundational theory, software, and methodology for building commitment-based distributed systems. To encourage wider adoption, the project will embed the algorithms in prototypes, and develop a tool-supported methodology for specifying and implementing social protocols. Further, Turtles will develop real systems based on use cases and practices from a number of industrial partners and evaluate these systems based on their feedback.

The success of Turtles will enable capturing important subtleties of real-life social and business interactions and transform how we design software systems for crucial multiparty domains such as healthcare, disaster response, smart cities, banking, education, and e-commerce and e-business, where commitments are crucial.

We list the potential beneficiaries and summarise the potential impact on each of them. (Please also see Pathways to Impact.)

KNOWLEDGE:

Researchers: Turtles will contribute foundational scientific knowledge crucial to building distributed systems of multiple parties. This will enable bringing together the currently distinct communities of programming languages and multiagent systems and lead to the formation of joint new communities and agendas based on computational social abstractions. The impact on service-oriented computing and business process management will also be profound.

Software developers: Turtles will contribute knowledge toward the development of systems that support business interactions within and across organisations. This knowledge will improve communication with stakeholders, and simplify the development and maintenance of such systems.

Stakeholders: Turtles will contribute knowledge that enables stakeholders to understand requirements and key performance indicators in terms of commitments, and to relate these artifacts to software.

PEOPLE:

Researchers and students: Turtles will train PDRAs and PhD students working on related projects. The project's results will also inform undergraduate and graduate students via courses and tutorials.

Software developers and stakeholders: Turtles will train software developers and stakeholders via the substantial engagement with our partners.

Protocol analysts: In the medium-to-long run, we envisage the emergence of a new class of professionals called protocol analysts, who will be experts in specifying, analysing, and implementing social protocols.

SOCIETAL:

Standards bodies: Standardisation is key to better interoperability and therefore better services in important societal domains such as healthcare, smart cities, energy, and so on. Standardisation efforts are enormously complex and standards are often not adopted. Sometimes, the reason for this is due to unclear and complex specifications. Turtles will simplify the creation of protocol-based standards via a formal, tool-supported language that enables expressing both the messaging and high-level concerns.

ECONOMIC:

Business ecosystems: Via easier standardisation, Turtles will support the proliferation of business ecosystems consisting of stakeholders, including government, and developers in many multiparty domains. This will lead to reduced implementation costs and increased business engagements via better interoperability. To give a concrete example, RosettaNet is used by Intel and other companies to conduct billions of dollars worth of business. A study showed that adopting the RosettaNet protocols reduced transaction costs and increased throughput (http://tinyurl.com/p38nc7r). A healthcare study postulated billions of dollars worth of savings for the United States both during protocol implementation and post-implementation periods (http://tinyurl.com/o7sv8y6).