Traditional BDI agents, rooted in logic programming, remain challenging to integrate with the hypermedia nature of open Web environments that rely on Semantic Web technologies like RDF and OWL. This paper examines the gap between these paradigms and surveys existing integration eforts on a conceptual and technical level and argues that existing tools ofer limited ergonomic support for developers. Our proposal for a deeper integration relies on a generalized BDI engine to enable the development of BDI agents that can directly reason and operate on semantic hypermedia resources. We derive ideal requirements and show an abstract architecture that can be tailored to support diferent types of beliefs and reasoning mechanisms.
Description Framework (RDF) [
In this paper, we reflect on these gaps, their technical implications, explore how this gap is currently being addressed in the hMAS community, and propose a mitigation strategy based on abstracting the design of agent-oriented programming languages and frameworks.
The BDI model [
To declaratively define the agent’s behavior the agent-oriented programming ( AOP) community has developed several BDI programming languages and frameworks [16]. Arguably, the most popular language is AgentSpeak(L) [17] with its Jason [18] implementation, which uses a formal syntax and semantics derived from LP for representing beliefs, goals, and actions as logic predicates.
Although BDI programming languages and frameworks with little or no reliance on LP mechanisms exist (e.g., [19, 20, 21]), they come with their own semantics and representation for beliefs, goals, and actions which still leave an abstraction gap to be covered with respect to Semantic Web technologies. In the remainder of this work we will refer to BDI agents implicitly assuming AgentSpeak(L)-like agents.
The World Wide Web was proposed by Tim Berners-Lee as a global, open information space based
on the ability to link resources through hypermedia documents [22]. The Web is built with the REST
architectural style [
The Semantic Web [23] complements the self-descriptive principle of REST by encoding semantics in
the representation through the general model of RDF [
Another fundamental principle is Hypermedia as the Engine of Application State (HATEOAS) which
defines hypermedia as “the simultaneous presentation of information and controls such that the
information becomes the afordance through which the user obtains choices and selects actions” [
The idea of hMAS [
Here we delve into the details of the gaps that we identified between BDI agents and semantic hypermedia, namely, (G1) the practical diferences for knowledge representation and manipulation between LP and Semantic Web technologies, and (G2) the limitations of BDI agents with open hypermedia environments, where actions are not predefined but discovered at runtime.
Syntactical Diferences Between LP and RDF. On a syntactical level, BDI agents commonly represent beliefs as definite clauses [ 28] (e.g. person(alice)). In the Semantic Web, knowledge is represented in the form of RDF triples which connect a subject to an object through a predicate (e.g. :alice rdf:type foaf:Person). Although RDF triples can be represented as definite clauses [ 29] (while the opposite is not true), in practice, semantic knowledge is available on the Web in RDF, and accessing that from BDI agents requires a translation step, which has many degrees of freedom (e.g., type(alice, person) and triple(alice, type, person) are reasonable mappings of the RDF triple above), which is completely up to BDI developers. In other words, implementing the translation is additional work which, if handled manually in an ad-hoc manner can be source of inconsistencies. Ontological Inference. RDF triplets are commonly paired with OWL ontologies (e.g., foaf2 in the example above), providing axioms that define the meaning of the triples’s predicates (e.g., rdf:type denotes the instance of relation between subjects and classes), as well as constraints, and relationships between concepts. This is a way to standardize knowledge representation across hypermedia. Logic programs, instead, are not really meant to be scattered over the Web, nor being accessed remotely.
Again, BDI developers who want to create agents capable of ontological inference face a hidden challenge: either they translate the OWL ontology into definite clauses, or they wrap some OWL reasoner into the agent interpreter and reasoning cycle. Despite being automatable to some extent (cf. [30]), both activities require additional engineering efort, and should rather be addressed with direct support from the BDI framework.
Querying. BDI commonly relies on Selective Linear Definite clause ( SLD) resolution [31] to query the belief base, when this is implemented as a logic program. Furthermore, they rely on logic unification [ 32] to match plans and goals, making the intertwining with LP even deeper. Conversely, the Semantic Web relies on SPARQL Protocol and RDF Query Language (SPARQL) [33] to query RDF data. SPARQL allows matching graph patterns and retrieve all the values of variables in the pattern from the matched triples, providing an intuitive way to query the knowledge base.
Despite both LP and SPARQL allow expressing declarative queries for a knowledge base, the latter may be more practical and eficient for RDF data, especially when the query spans over multiple triples. Consider for instance a query searching for the unknown relations ?r between alice and bob, and matching all the triples where the subject is carl and the predicate is ?r to retrieve all the objects ?x. A query of this kind would be straightforward to write in SPARQL, while it would require second-order variables in LP. This feature is not commonly supported by LP or BDI frameworks, and is typically worked around via meta-programming, resulting in longer and more cumbersome LP code.
As for inference, given the expressiveness limitations of OWL, Semantic Web Rule Language (SWRL) [34] can be used to define inference rules. Here, a similar argument like the one above concerning the wrapping of SWRL reasoners applies.
Open vs. Close Environments. BDI agents are equipped with a predefined set of plans, composed of sequences of actions that the agent knows how to execute. Additionally, BDI agents usually run in a closed environment, which is designed by the developer to provide perceptions to the agents and support the execution of the actions used in the plans [25]. The development process of BDI agents is hence based on the fact that a developer anticipates the relevant possible settings an agent may encounter and provides plans to handle them. This makes it dificult for agents to adapt to unexpected settings or to flexibly incorporate new actions that become available while exploring the environment. Relating this to the open hypermedia nature of the Web, makes the efective integration of BDI agents challenging. Following the HATEOAS principle, agents may discover new afordances (i.e., possible actions) while navigating through the hypermedia environment. The inability to exploit such new afordances may limit the agent in adapting to the environment and efectively interact with it.
Guided by our gap analysis, we identify a set of requirements that we believe may improve practical development of BDI agents in Web environments. We hence propose that a BDI agent programming framework for hMAS should support the following features: (R1) direct manipulation of RDF and OWL triples in the agent’s belief base; (R2) ontological inference for deliberation (e.g., plan selection and execution); (R3) support for querying the belief base via SPARQL; (R4) ability to exploit afordances discovered in the environment to dynamically adapt the agents’ plans.
Here we summarize some related works demonstrating how, despite the many extensions of AgentSpeak towards DL, the direct exploitation of Semantic Web technologies (namely, (R1) and (R3)) in BDI programming frameworks is still underexplored.
In [35] authors propose AgentSpeak-DL, an extension of AgentSpeak to support the ℒ description logic [36]. There, agents’ belief bases consist of an ℒ ontology (concepts, roles, and individuals) and any operation involving access to the belief base (e.g., plan selection, belief update, and query) imply either performing inference or updating such ontology. So, AgentSpeak-DL provides the theoretical foundations to support our integration requirements from Section 3.1. AgentSpeak-DL is then extended by CooL-AgentSpeak [37], where authors explore inter-agent exchange of procedural knowledge [38], possibly expressed by diferent ontologies. The authors of [ 39] provide a diferent theoretical integration of AgentSpeak with DL to account for goal states (new, suspended, active, succeeded, failed) in the agent’s lifecycle, in such a way that state transitions are tied to conditions expressed in DL.
AgentSpeak-DL also serves as the foundation for JASDL [40], an implementation of AgentSpeak-DL in Jason [18]. The key extension in JASDL is the introduction of “semantically-enriched literals,” i.e., ℒ statements expressed as logic facts, (see example in Listing 1). JASDL adopts Jason belief syntax and Listing 1: On the left, RDF triples expressed as predicates in JASDL, on the right their meaning inferred by a
OWL reasoner. The example is taken from [40]. hotel(hilton)[o(travel)]. hasRating(hilton, threeStarRating)[o(travel)]. city(london)[o(travel)].
// hilton is a hotel // hilton has three-star rating // london is a city // plain RDF form targetEquipment(Machine) :rdf("https://territoire.emse.fr/kg/emse/fayol", "https://w3id.org/bot#containsElement", Machine) & rdf(Machine, "http://www.w3.org/1999/02/22-rdf-syntax-ns#type", "http://www.productontology.org/id/Filler_( packaging)") . // equivalent short form based on OntologyArtifact properties targetEquipment(Machine) :- containsElement(fayol, Machine) & ’Filler_(packaging)’(Machine). Listing 2: The example shows how RDF triples are represented in Hypermedea agents belief base. The example is taken from Hypermedea Application Programming Interface (API) reference (https://archive.is/vxULu). belief-base management mechanisms, while wrapping and adapting a semantic reasoner to let agents perform ontological inference, thus addressing (R2). With this approach, developers gain ontological inference in Jason, but at the price of translating RDF triples into logic predicates.
Other approaches [37, 41, 42, 43] rely on encapsulating functionalities and services in external components that agents can share and exploit to support their activities. For instance CooL-AgentSpeak [37] relies on CArtAgO [44], to define and implement the Ontology Artifact, providing agents with ontological inference capabilities—encoding DL statements as beliefs similarly to JASDL.
Similarly, Hypermedea [41] extends JaCaMo [45] (again, via artifacts and constraints on the beliefsbase syntax) to support agents interacting with semantic hypermedia and, in particular, the WoT. More precisely, Hypermedea artifacts produce agent beliefs such as the ones in Listing 2 through observable properties—i.e., RDF triples expressed as logic predicates, in Jason’s syntax. Furthermore, other artifacts support the manipulation of such RDF triples, materialization of ontological inferences, as well as the interaction with WoT’s things, and the synthesis of plans via a PDDL planner—hence addressing (R4).
A diferent take is proposed in [ 42] where Astra [46] agents are equipped with a set of modules to allow BDI agents to store knowledge and interact with a “personal” triple store separated from the belief base. Similarly to other approaches, the bridge to belief representation is by mapping triples directly to logic predicates with three arguments.
Finally, although not directly related to BDI agents, the Yggdrasil framework [24] provides a hypermedia layer for building hMAS environments on top of CArtAgO artifacts and has been used to implement hMAS applications with the JaCaMo framework.
Exploiting Discovered Actions at Runtime. To exploit new actions discovered at runtime (R4) BDI agents must have access to a planner either within the agent (see [47] for a general overview of planning in BDI agents) or through the environment—as proposed in Hypermedea [41]. In the context of hMAS, this requires that the environment formally describes how and when to use them (e.g., their preconditions and efects), and that agents can interpret such information to instruct the planner accordingly. Recent works explored this idea by using signifiers [ 48] in hypermedia environments to better convey afordances to BDI agents [27]. A less structured approach is proposed in [49, 50], where agents (albeit not BDI) are able to select actions in a hypermedia environment by using a LLM oracle.
To bridge the gap between BDI agents and semantic hypermedia, we propose a generalized BDI engine, addressing the requirements in Section 3.1. This approach primarily tackles gap (G1), while leaving the exploration of gap (G2) (benefitting from the modularity of the engine) to future works. A similar approach has been proposed in [51], in which authors discuss the idea of a modular BDI architecture, to make it independent of the logic representation and reasoning mechanisms, although not in the context of hypermedia. Sharing the core modularity idea, we propose to create a BDI interpreter encapsulating agents reasoning cycle, while decoupling the specific technology used for knowledge representation and manipulation. This would allow for tailoring the reasoning mechanisms to application-specific needs. In this way, beliefs can flexibly support diferent formats depending on the domain at hand and be adapted to work directly with e.g., RDF triples, to JSON, YAML, and other Web standards. Such a design fulfills the first requirement (R1).
However, separating the reasoning cycle of BDI agents from knowledge representation is complex, as the two are deeply intertwined [35]. Consider for instance these fundamental Figure 1: Generalized Architecture. operations: (i) plan selection depending on conditions to be Generalized BDI Reasoning Model tested against the belief base; (ii) belief querying as part of a Agent Lifecycle sopoanlfakanleco,tgofioiofcrnuut.phnIdeniaficstLaaitPnkio-egbnotahfaseedndedacgfrirseeainsomotn’les-umwtkioaonkrnoki,wnswg,le;thhd(eigeirisee)e,absawesilnoipeufaalurdgtpesdoniafmetarepacfllyooizruretredhslyee GAenbestrricacBtieolnieUsfes AbQsFtuurnaeccrUttysieoiBnsneglief AbsFUturpnadcctUtaisoBteesnelief AbSFsuteUrslnaeesccctttiiooPnnlan BDI engine, each operation may rely on diferent matching Defines strategies. For instance, for Semantic Hypermedia agent, one Defines Concrete Belief Defines Defines fmoarybreelileyfoqnuRerDyFinagndanOdWuLpdfoarteb,ealniedfsSrWepRreLsefonrtaptliaonn,sSePleAcRtiQonL CRoenpcreresteentBaetiloienf QFuunecrtyioinng ConFUcurpnedctetaiotBenelief ConScerleectetioPnlan and ontological inference—thus addressing (R2) and (R3). Concrete Specification Function
The generalized BDI engine can be realized with a layered Uses architecture (see Figure 1) comprising three layers. Layer 1: generalized BDI reasoning layer, where the agent’s de- Application liberation process is implemented as a reasoning cycle, yet agnostic to how beliefs are represented and manipulated. This is where agents’ intentions are maintained and updated as agents perceive, decide what to do, and ifnally act. Any operation involving belief bases should rely on some abstract API, to be implemented by the next layer. Layer 2: concrete specification layer, where the aforementioned API is implemented to provide concrete operations on the belief base. This layer acts as an adapter between the generalized BDI reasoner and the specific knowledge representation technologies of choice. Lastly, Layer 3: application layer, where actual goals, plans, and interactions of a MAS are defined to tackle some specific use case.
Requirement (R4) although not directly addressed by the approach, could be integrated with the techniques already being explored in the hMAS community (cf. Section 4), benefitting from the possibility of choosing the most suitable knowledge representation technology to facilitate their agents to discover new actions at runtime in the hypermedia environment.
In this paper, we present a gap analysis between BDI agents and semantic hypermedia, showing how they are dificult to integrate due to the diferent technologies and paradigms they rely on. We identify a set of ideal requirements, finding that none of the surveyed related works fully satisfy them and propose to implement them through a generalized BDI engine that can be tailored to use RDF and OWL as knowledge representation technologies.
Future work will focus on implementing the architecture proposed in Section 5 and explore how to support hMAS developers with syntax extensions and tools to ease the development of BDI agents in hypermedia environments within our BDI programming framework JaKtA [52].
This work has been partially supported by: (i) “WOOD4.0 - Woodworking Machines for Industry 4.0”, Emilia-Romagna CUP E69J22007520009; (ii) “FAIR–Future Artificial Intelligence Research”, Spoke 8 “Pervasive AI” (PNRR, M4C2, Investimento 1.3, Partenariato Esteso PE00000013), funded by the EC under the NextGenerationEU programme; (iii) “ENGINES — ENGineering INtElligent Systems around intelligent agent technologies” project funded by the Italian MUR program “PRIN 2022” (G.A. 20229ZXBZM), and (iv) 2023 PhD scholarship (PNRR M4C2, Investimento 3.3 DM 352/2022), co-funded by the European Commission and AUSL della Romagna.
During the preparation of this work, the authors used GitHub’s Copilot in order to spell-check their writing, and make it more concise. After using this tool, the authors reviewed and edited the content as needed and take full responsibility for the content of the publication. [14] B. Motik, P. F. Patel-Schneider, B. Parsia, OWL 2 Web Ontology Language Structural Specification and Functional-Style Syntax (Second Edition), W3C Recommendation 11 December 2012, W3C Recommendation, World Wide Web Consortium, 2012. URL: http://www.w3.org/TR/2012/ REC-owl2-syntax-20121211/. [15] M. P. Georgef, A. L. Lansky, Procedural knowledge, Proc. IEEE 74 (1986) 1383–1398. doi: 10.
1109/PROC.1986.13639. [16] V. Mascardi, D. Demergasso, D. Ancona, Languages for programming BDI-style agents: an overview, in: F. Corradini, F. D. Paoli, E. Merelli, A. Omicini (Eds.), WOA 2005 - Workshop "From Objects to Agents", Pitagora Editrice Bologna, 2005, pp. 9–15. URL: http://lia.deis.unibo.it/books/ woa2005/papers/2.pdf. [17] A. S. Rao, AgentSpeak(L): BDI agents speak out in a logical computable language, in: W. V. de Velde, J. W. Perram (Eds.), Agents Breaking Away, 7th European Workshop on Modelling Autonomous Agents in a Multi-Agent World, volume 1038 of Lecture Notes in Computer Science, Springer, 1996, pp. 42–55. doi:10.1007/BFB0031845. [18] R. H. Bordini, J. F. Hübner, M. Wooldridge, Programming multi-agent systems in AgentSpeak using Jason, John Wiley & Sons, 2007. [19] A. Pokahr, L. Braubach, W. Lamersdorf, Jadex: A BDI reasoning engine, in: R. H. Bordini, M. Dastani, J. Dix, A. E. F. Seghrouchni (Eds.), Multi-Agent Programming: Languages, Platforms and Applications, volume 15 of Multiagent Systems, Artificial Societies, and Simulated Organizations, Springer, 2005, pp. 149–174. [20] T. Kampik, J. C. Nieves, JS-son - A lean, extensible JavaScript agent programming library, in: L. A. Dennis, R. H. Bordini, Y. Lespérance (Eds.), Engineering Multi-Agent Systems (EMAS) 7th International Workshop, volume 12058 of Lecture Notes in Computer Science, Springer, 2019, pp. 215–234. doi:10.1007/978-3-030-51417-4_11. [21] N. Howden, R. Rönnquist, A. Hodgson, A. Lucas, JACK intelligent agents-summary of an agent infrastructure, in: 5th International conference on autonomous agents, volume 6, 2001, p. 40. [22] T. Berners-Lee, R. Cailliau, J. Grof, The World-Wide Web, Comput. Networks ISDN Syst. 25 (1992) 454–459. doi:10.1016/0169-7552(92)90039-S. [23] T. Berners-Lee, J. Hendler, O. Lassila, The semantic web, Scientific American 284 (2001) 34–43. URL: http://dx.doi.org/10.1038/scientificamerican0501-34. doi: 10.1038/ scientificamerican0501-34. [24] A. Ciortea, O. Boissier, A. Ricci, Engineering World-Wide multi-agent systems with hypermedia, in: D. Weyns, V. Mascardi, A. Ricci (Eds.), Engineering Multi-Agent Systems (EMAS) 6th International Workshop, volume 11375 of Lecture Notes in Computer Science, Springer, 2018, pp. 285–301. doi:10. 1007/978-3-030-25693-7_15. [25] D. Weyns, A. Omicini, J. Odell, Environment as a first class abstraction in multiagent systems,
Auton. Agents Multi Agent Syst. 14 (2007) 5–30. doi:10.1007/S10458-006-0012-0. [26] A. Ricci, M. Piunti, M. Viroli, Environment programming in multi-agent systems: an artifact-based perspective, Auton. Agents Multi Agent Syst. 23 (2011) 158–192. doi:10.1007/ S10458-010-9140-7. [27] D. Vachtsevanou, et al., Enabling BDI agents to reason on a dynamic action repertoire in hypermedia environments, in: M. Dastani, et al. (Eds.), Proceedings of the 23rd International Conference on Autonomous Agents and Multiagent Systems (AAMAS), ACM, 2024, pp. 1856–1864. doi:10.5555/ 3635637.3663048. [28] J. Makowsky, Why horn formulas matter in computer science: Initial structures and generic examples, Journal of Computer and System Sciences 34 (1987) 266–292. doi:https://doi.org/ 10.1016/0022-0000(87)90027-4. [29] B. N. Grosof, I. Horrocks, R. Volz, S. Decker, Description logic programs: Combining logic programs with description logic, in: G. Hencsey, et al. (Eds.), Proceedings of the Twelfth International World Wide Web Conference (WWW), ACM, 2003, pp. 48–57. doi:10.1145/775152.775160. [30] K. Samuel, et al., Translating OWL and semantic web rules into Prolog: Moving toward description logic programs, Theory and Practice of Logic Programming 8 (2008) 301–322. doi:10.1017/ S1471068407003249. [31] M. H. van Emden, R. A. Kowalski, The semantics of predicate logic as a programming language, J.
ACM 23 (1976) 733–742. doi:10.1145/321978.321991. [32] A. Martelli, U. Montanari, An eficient unification algorithm, ACM Trans. Program. Lang. Syst. 4 (1982) 258–282. doi:10.1145/357162.357169. [33] Feigenbaum, Williams, Clark, Torres, SPARQL 1.1 Protocol, W3C Recommendation 21 March 2013, 2013. URL: http://www.w3.org/TR/2013/REC-sparql11-protocol-20130321/. [34] I. Horrocks, P. F. Patel-Schneider, H. Boley, S. Tabet, B. Grosof, M. Dean, et al., SWRL: A semantic web rule language combining OWL and RuleML, Technical Report 79, 2004. [35] Á. F. Moreira, R. Vieira, R. H. Bordini, J. F. Hübner, Agent-oriented programming with underlying ontological reasoning, in: M. Baldoni, U. Endriss, A. Omicini, P. Torroni (Eds.), Declarative Agent Languages and Technologies (DALT) Third International Workshop, volume 3904 of Lecture Notes in Computer Science, Springer, 2005, pp. 155–170. doi:10.1007/11691792_10. [36] F. Baader, I. Horrocks, U. Sattler, Chapter 3 Description Logics, in: F. van Harmelen, V. Lifschitz, B. Porter (Eds.), Handbook of Knowledge Representation, volume 3 of Foundations of Artificial Intelligence, Elsevier, 2008, pp. 135–179. doi:10.1016/S1574-6526(07)03003-9. [37] V. Mascardi, D. Ancona, M. Barbieri, R. H. Bordini, A. Ricci, CooL-AgentSpeak: Endowing AgentSpeak-DL agents with plan exchange and ontology services, Web Intelligence and Agent Systems 12 (2014) 83–107. doi:10.3233/WIA-140287. [38] D. Ancona, V. Mascardi, Coo-BDI: Extending the BDI model with cooperativity, in: J. A. Leite, A. Omicini, L. Sterling, P. Torroni (Eds.), Declarative Agent Languages and Technologies (DALT) First International Workshop, volume 2990 of Lecture Notes in Computer Science, Springer, 2003, pp. 109–134. doi:10.1007/978-3-540-25932-9_7. [39] T. G. Halaç, E. E. Ekinci, O. Dikenelli, Description logic based BDI implementation for goaldirected semantic agents, in: O. Boissier, et al. (Eds.), Proceedings of the 2011 IEEE/WIC/ACM International Conference on Intelligent Agent Technology (IAT), IEEE Computer Society, 2011, pp. 62–65. doi:10.1109/WI-IAT.2011.192. [40] T. Klapiscak, R. H. Bordini, JASDL: A practical programming approach combining agent and semantic web technologies, in: M. Baldoni, T. C. Son, M. B. van Riemsdijk, M. Winikof (Eds.), Declarative Agent Languages and Technologies (DALT) 6th International Workshop, volume 5397 of Lecture Notes in Computer Science, Springer, 2008, pp. 91–110. doi:10.1007/978-3-540-93920-7_7. [41] V. Charpenay, A. Zimmermann, M. Lefrançois, O. Boissier, Hypermedea: A framework for Web (of Things) agents, in: F. Laforest, R. Troncy, E. Simperl, D. Agarwal, A. Gionis, I. Herman, L. Médini (Eds.), Companion of The Web Conference, ACM, 2022, pp. 176–179. doi:10.1145/3487553. 3524243. [42] E. O’Neill, K. Beaumont, N. V. Bermeo, R. W. Collier, Building management using the semantic web and hypermedia agents, in: T. Käfer, A. Harth, A. Ciortea, V. Charpenay (Eds.), Proceedings of the All the Agents Challenge (ATAC), volume 3111 of CEUR Workshop Proceedings, CEUR-WS.org, 2021, pp. 32–37. URL: https://ceur-ws.org/Vol-3111/short5.pdf. [43] A. Ciortea, O. Boissier, A. Ricci, Engineering World-Wide multi-agent systems with hypermedia, in: D. Weyns, V. Mascardi, A. Ricci (Eds.), Engineering Multi-Agent Systems (EMAS) 6th International Workshop, volume 11375 of Lecture Notes in Computer Science, Springer, 2018, pp. 285–301. doi:10. 1007/978-3-030-25693-7_15. [44] A. Ricci, M. Piunti, M. Viroli, A. Omicini, Environment programming in CArtAgO, in: R. H.
Bordini, M. Dastani, J. Dix, A. E. F. Seghrouchni (Eds.), Multi-Agent Programming, Languages, Tools and Applications, Springer, 2009, pp. 259–288. doi:10.1007/978-0-387-89299-3_8. [45] O. Boissier, R. H. Bordini, J. F. Hübner, A. Ricci, A. Santi, Multi-agent oriented programming with JaCaMo, Science of Computer Programming 78 (2013) 747–761. doi:10.1016/j.scico.2011. 10.004. [46] R. W. Collier, S. E. Russell, D. Lillis, Reflecting on agent programming with AgentSpeak(L), in: Q. Chen, P. Torroni, S. Villata, J. Y. Hsu, A. Omicini (Eds.), Principles and Practice of Multi-Agent Systems (PRIMA) 18th International Conference, volume 9387 of Lecture Notes in Computer Science, Springer, 2015, pp. 351–366. doi:10.1007/978-3-319-25524-8_22. [47] F. Meneguzzi, L. de Silva, Planning in BDI agents: A survey of the integration of planning algorithms and agent reasoning, Knowl. Eng. Rev. 30 (2015) 1–44. doi:10.1017/S0269888913000337. [48] D. Vachtsevanou, A. Ciortea, S. Mayer, J. Lemée, Signifiers as a first-class abstraction in hypermedia multi-agent systems, in: N. Agmon, B. An, A. Ricci, W. Yeoh (Eds.), Proceedings of the 2023 International Conference on Autonomous Agents and Multiagent Systems (AAMAS), ACM, 2023, pp. 1200–1208. doi:10.5555/3545946.3598763. [49] D. Vachtsevanou, J. Lemee, R. Rot, S. Mayer, A. Ciortea, G. Ramanathan, HyperBrain: Humaninspired hypermedia guidance using a large language model, in: Proceedings of the 34th ACM Conference on Hypertext and Social Media, HT ’23, Association for Computing Machinery, New York, NY, USA, 2023. doi:10.1145/3603163.3609077. [50] S. Schmid, M. Freund, A. Harth, Adaptive planning on the web: Using LLMs and afordances for web agents, in: S. Tiwari, B. Villazón-Terrazas, F. Ortiz-Rodríguez, S. Sahri (Eds.), Knowledge Graphs and Semantic Web - 6th International Conference (KGSWC), volume 15459 of Lecture Notes in Computer Science, Springer, 2024, pp. 93–108. doi:10.1007/978-3-031-81221-7_7. [51] P. Novák, J. Dix, Modular BDI architecture, in: H. Nakashima, M. P. Wellman, G. Weiss, P. Stone (Eds.), 5th International Joint Conference on Autonomous Agents and Multiagent Systems (AAMAS), ACM, 2006, pp. 1009–1015. doi:10.1145/1160633.1160814. [52] M. Baiardi, S. Burattini, G. Ciatto, D. Pianini, Blending BDI agents with object-oriented and functional programming with JaKtA, SN Comput. Sci. 5 (2024) 1003. doi:10.1007/ S42979-024-03244-Y.