One research focus is the development of mechanisms that allow an accessible representation of institutional concepts, both for modellers and domain experts. These are highlighted informally on this page. For a more comprehensive treatment (and corresponding academic literature), please refer to the associated publications listed at the bottom of this page or under publications.

As part of this we have developed Nested ADICO (nADICO), a refined variant of Crawford and Ostrom’s Grammar of Institutions (Crawford and Ostrom, 1995) to allow agents to infer complex institutional understanding (specifically normative behaviour) at runtime.


The original ADICO grammar represents an idealised representation of institutions but providing a generic grammar to express various institution types, including shared strategies (conventions), norms and rules. The ADICO components include acting parties (Attributes), a Deontic, and aIm, Conditions, and potential consequences (Or else). Combining specific components allows the representation of institutional statements, which represent specific institution types such as conventions (Crawford and Ostrom refer to those as shared strategies), norms and rules. Using a stylised scenario of corruption, let’s exemplify the different institution types (institutional grammar components that constitute a specific institution type are highlighted in parentheses):

  • Shared strategy/Convention (AIC) –  ‘Citizens (A) do not offer (I) bribes under any circumstance (C).’;
  • Norm (ADIC) – ‘Citizens (A) must not (D) offer bribes (I) under any circumstance (C).’;
  • Rule (ADICO) – ‘Citizens (A) must not (D) offer bribes to officials (I) under any circumstance (C), or else the officials must notify the police (O).’

Note: The ‘Conditions’ component C does not need to be explicitly specified if the institutional statement applies to any circumstance.


nADICO takes the representation a step further by a) allowing the existence for consequences for norms (as opposed to only rules), and b) enabling a more comprehensive representation of complexity (e.g., multiple consequences for norm violations; horizontal and vertical nesting – see below). nADICO’s underlying principles include:

  • Norm Emergence: Individuals learn injunctive norms (norms that characterise prohibitions and obligations) based on the associated sanctions (e.g., by experience, observation, or communication). The fundamental nADICO characteristic that enables this is the consideration of sanctions for norms.
  • Comprehensive Representation: Individuals can infer more complex normative understanding by inferring nested interdependencies of institutions, which is particularly important for a comprehensive computational representation (e.g. to allow agents to explain why they adopt a specific institution understanding, to infer sanctions for norm violations, etc.). The fundamental conceptual modification in nADICO is the substitution of the ‘Or else’ component with an nADICO statement (i.e. another convention, norm, or rule), which enables the nesting of statements.
  • Endogenous Perspective: nADICO focuses on an endogenous dynamic institution perspective (especially in conjunction with the concept of Dynamic Deontics), whereas the original grammar is focused on a static analysis of observed institutions, generally from a descriptive perspective (e.g., because of lacking insight of the underlying motivations for norms, etc.). While a normative background can be injected into agents (e.g. as initial set of norms or rules), the aforementioned modifications enable the runtime identification of norms.

Let’s discuss the implications of the modifications by continuing our previous example. Using nADICO individuals could infer that

‘Citizens (A) must not (D) offer bribes (I) under any circumstance (C), or else officials (A) may (D) educate them (I) if the citizen is unaware of the local practice (C), or officials (A) may (I) report them to the police (I).’

Extracting the grammar components, this norm can now be comprehensively expressed as ADIC(ADIC or ADIC). Of course, this is just a simple example, but norms could also include co-occurring consequences (e.g. the bribing citizen being educated and reported to the police: ADIC(ADIC and ADIC)), or allow the specification of consequences for the official if (s)he does not comply (e.g., a supervisor reporting on a violating official: ADIC(ADIC and ADIC)ADIC). (For more details, examples, formal syntax, as well as differentiation between norms and rules, please refer to the associated publications.)

Horizontal and Vertical Nesting

The introduction of nested statements allows the representation of complexity on a specific institutional level (e.g. multiple co-occurring actions) – referred to as horizontal nesting – , but also interdependencies across multiple levels (structural institutional regress) – which we refer to as vertical nesting. The following schema illustrates these characteristics. Statements on a given institutional level (monitored statements) are guarded by consequential statements on the next higher level that prescribe consequences for violating monitored statements (Recall the citizen’s behaviour as a monitored statement, and the official’s behaviour as a consequential statement). Using this approach, statements can be specified with an arbitrary level of complexity.

Associated publications:

C. K. Frantz, M. K. Purvis, B. T. R. Savarimuthu, M. Nowostawski: Modelling Dynamic Normative Understanding in Agent Societies. Scalable Computing: Practice and Experience, vol. 16, no. 4, pp. 355-378, 2015 [PDF]

C. K. Frantz: Agent-Based Institutional Modelling: Novel Techniques for Deriving Structure from Behaviour, PhD thesis, University of Otago, 2015, Available under: [PDF]

C. Frantz, M. K. Purvis, M. Nowostawski, B. T. R. Savarimuthu: nADICO: A Nested Grammar of Institutions. PRIMA 2013: Principles and Practice of Multi-Agent Systems, Springer, LNCS 8291, 2013, pp. 429-436 [PDF]


Crawford S. and Ostrom E. (1995): A grammar of institutions. American Political Science Review 89(3): 582600