i* is a requirements modelling language that is founded on social concepts (actors, social dependencies). Azzurra is a business process speci cation language that uses roles and commitments as primitives. Symboleo is a smart contract speci cation language grounded in legal concepts (roles, parties, obligations, powers). We compare and contrast the social dependence relationships used by the three languages.
One of the key elements of i* [
Given that the languages target di erent domains, di erent examples are used for illustration. Note also that a full comparison of these languages is beyond the scope of this paper. In an i* social dependency (Fig. 1), the depender wants something and the dependee is able and willing to deliver it. However, the force of the dependence can
Copyright © 2020 for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0). vary from weak to strong on either side of the dependum. Consider a customer's dependence on a shop to nd her favourite hair shampoo: it is weak because there are probably other stores that sell it as well, and other shampoos are also comparable. Contrast this with one's dependence on a renowned surgeon for a rare medical operation. This one is strong, because there are no substitute dependees. The force of the dependence on the dependee's side is even more important, as she is responsible for delivering on the dependence. That force is de ned along two dimensions: ability to deliver on the dependum and degree of commitment to deliver. The dependence on the renowned surgeon is strong on ability and medium on commitment, as surgeons will postpone operations in case of an emergency.
Now, consider a beggar who depends on passersby to ful ll her goal of having some money. This is a social dependence too and it is weak on both sides: the beggar can switch to another kind of dependence to get some money, e.g, work, while the passersby have not even agreed explicitly to the beggar to deliver on the dependum, they are just willing to do it, occasionally. Here, the dependence is established statistically: some passersby give money, and sooner or later this establishes a dependency for the beggar.
i* does recognize the importance of the force of a dependence by allowing
three possible levels of force: critical, committed, and open [
i* was developed at the University of Toronto in the early `90s as part of
Eric Yu's PhD thesis, with collaborators Lawrence Chung, Brian Nixon, and
John Mylopoulos. It further led to the development of many other languages
and dialects, including the Goal-oriented Requirement Language, part of the
User Requirements Notation standard, with collaborators documented in [
Originating in the area of multi-agent systems [
Azzurra is a conceptual modelling language for business processes. Its main thesis is that such processes, being social artifacts, need to be de ned in social terms, rather than system-oriented ones (e.g., Petri nets or BPMN). Accordingly, business processes (aka `protocols') are de ned in terms of roles and commitments, with constraints attached. Azzurra models can be seen as requirements speci cations for business processes, as they describe what a business process is supposed to achieve without getting into the details of how to achieve it.
Figure 2 (adapted from [
There are nine commitments for this protocol, each using a <trigger> <commitment> format. The rst, C1, is triggered when the protocol is instantiated, has as roles the specialist (debtor) and the patient (creditor), is unconditional (antecedent= true), and is ful lled when the patient is i) examined, then ii) diagnosed, and then iii) dis-hospitalized. The second commitment, C2, is triggered if there is no need for X-rays and is ful lled when a sling is made for the patient. Protocol re nements constrain the agents that participate in a protocol instance. For example, agents may be constrained on how many concurrent commitments they have for a given role, such as a surgeon for treatment protocol instances. Finally, a knowledge base de nes some domain axioms, which can be used to reason about propositions used as triggers or as elements of commitment antecedents and consequents.
Azzurra supports two types of reasoning. ENACTPROTOCOL determines how an event updates the state of a protocol instance and of the commitment instances therein. CHECKCOMPLIANCE checks whether an occurred event violates the speci cation of a protocol instance. This corresponds to identifying commitments that are not created/ful lled, unexpected commitment operations, and protocol constraint violations.
In summary, commitments specialize and improve the formalization of social dependence relationships. They also come with a richer language than i* for modeling business processes in an outcome-oriented way. Azzurra was developed at the University of Trento between 2012 and 2016 and only managed a few publications including a best paper, awarded at RCIS 2015. The collaborators for that project included Fabiano Dalpiaz, Evellin Cardozo, Giulia Canobbio, Paolo Giorgini, and John Mylopoulos. 4
Obligations are commitments with legal force. The legal force is de ned through powers that a creditor has towards the debtor of an obligation to cancel or suspend an obligation or another power, or to initiate new obligations or powers. Obligations constitute a specialization of the concept of commitment in that they can be created, cancelled, etc. by someone who has the power to do so. In turn, powers constitute specializations of obligations in that they can include in their consequent the creation, cancellation, etc. of other powers or obligations.
Obligations and powers can be found in legal contracts. Legal contracts can be thought as processes too, but they are legal rather than business ones, in that they describe the space of allowable executions that comply with terms and conditions of a contract. The presence of powers in legal contracts make them a much more malleable concept than that of a business process in that they can be reshaped through powers with the introduction/cancellation of obligations while a contract is being executed (i.e., \performed" in Law).
Symboleo is a formal speci cation language for legal contracts, intended to formalize requirements for smart contracts. These are software systems running on a blockchain platform (or involving a regular database) that monitor and control the execution of a legal contract. Symboleo is founded on an ontology for contracts that is centered around the notions of obligation and power, and includes role and party (the actors playing roles in a contract), asset, as well as situation and event. Situations occur over time, e.g., the situation of commuting to work. Events, on the other hand, happen instantaneously, e.g., arrivedAtWork.
The language adopts many elements from Azzurra. Obligations and powers use the same format as commitments. Their antecedents, consequents, and triggers are expressed as events happening in a certain order and satisfying constraints. For example, for a sale contract, the consequent of a delivery obligation may be \Sale item delivered to delivery address by delivery date".
Table 1 presents a Symboleo speci cation for a contract (adapted from [
The core of a contract speci cation are its obligations and powers. In our example there are two obligations: the seller must deliver the sale item to the delivery address by the delivery date (O1), while the buyer must pay on time the sale amount (O2). The contract also includes one power: if the buyer violates the payment obligation, the seller has the power to terminate the contract (P1). Note that the creditor of a power may choose to not exercise it.
Legal contracts are complex constructs with many features that go well beyond those of business processes. For instance, some obligations may apply after the successful termination of a contract (and are accordingly called surviving obligations), e.g., a con dentiality obligation for a sale transaction may apply 6 months after a contract terminates. Contracts may spawn subcontracts that may be established while a contract is executing. For example, when a developer undertakes a large project in the construction industry, she may not have lined up all the subcontractors and their respective subcontracts.
Symboleo speci cations can be validated to ensure that they are consistent with the expectations of the contracting parties by a tool (https://doi.org/10. 5281/zenodo.3903954) that enacts scenarios and determines the contract nal state. For example, for the scenario \Seller delivers on time, buyer does not pay on time, seller exercises power to terminate", the tool determines that the nal state of the contract is `cancelled'. The scenarios for validation are provided by contracting parties, along with their anticipated nal state of the contract when each scenario is enacted.
The Symboleo project started in January 2018 and is ongoing at the University of Ottawa with the co-authors of this paper as main contributors. 5
The concept of social dependence relationships in RE was pioneered by i*. This provided a solid conceptual base that is being proven to be of lasting value for developments in other application domains. With changes in syntax and semantics, it led to formalizing the concept of commitment in business processes in Azzurra. Then came the introduction of contractual obligations and powers in Symboleo, which leads to the possibility of monitoring legal compliance in contracts, possibly with blockchain support. We expect the social dependence concept to in uence other languages in the future.