Problem Statement

The design process exhibits a certain friction in its digital communicative processes that is holding back the AEC industry from fully capitalising on the digital revolution.

Read More: Chapter 1 Introduction

The original research question, as set out by the InnoChain call, was to analyse how complex digitally based design can be communicated and collated internally, within a design team, and externally, with the various stakeholders involved in the design process. I have broken this down into three more specific questions that each tackle a critical aspect of digital communication in the AEC industry:

Literature Review

This chapter is probably my favourite one. In it, I am establishing a theoretical base around communication in AEC and its relationship with the wickedness of design problems.

The most important takeaway is that communication is an integrated technical process and social phenomena. The social, inferential aspects of it have been neglected in recent research and discourse—or coerced into patterns shaped primarily by metaphors derived from its technical facets.

Read More: Chapter 2 Literature Review


The process of modern societal & scientific differentiation explains the resulting ontological divergence of disciplines (& stakeholders) in the design process.

2.2 Differentiation and Divergence


Design problems are, nevertheless, a convergent force: productive multi-stakeholder interaction leads to the emergence of shared understanding through iterative ontological displacement and dialogue.

2.3 Design & Convergence


Communication is underpinned by communicative contracts between people - not machines. Technical models of communication describe a machine-to-machine process, whereas inferential models capture the social aspects of communication.

State of the art

Prototypical digital design information systems have evolved from research into how to best manage wicked design problems through data-informed dialogue. With the advent of BIM, research & discourse emphasised technical concerns at the expense of the social aspect of design communication.

Methodology & Speckle

My research unfolded in the context of a living laboratory. In late 2015, I started developing Speckle as a means to contrast current approaches to new ones. Speckle also served (and keeps serving, but that's a different story!) as a vehicle to gather both quantitative and qualitative data in real world scenarios—not just artificial case studies.

Read More: Chapter 3 Methodology

Data Representation

Ontological Revision; Composability vs. Completeness

In this chapter, I am primarily investigating whether sketching with digital representations of design data is beneficial for the design process. To do so, I am comparing Speckle's self-describing, composable object model strategy against an ontologically complete “one standard” approach.

Read More: Chapter 4 Data Representation

Key Sections

4.1 Composable Data Structures 4.2 Encoding Existing Ontologies 4.4 Managing Ontological Diversity

Does digital ontological revision happen?

Yes; a composable object model enables a productive process of “sketching with data”: end-users do create their own “on the fly” ontologies (18% of total objects are user-defined).

Does this result in a more efficient representation of data?

End-user driven ontological revision produces smaller (in both complexity, as measured by tree depth, and actual size) data structures as opposed to fixed, pre-defined ontologies (IFC).

Data relevancy

This validates the maxim of relevance, whereby in a communicative exchange, information tends to a maximally relevant state (as defined by the lowest possible cognitive cost and the maximum potential impact).

Data Classification

Centralised, File Based Classification vs. Curated, Object Based System.

In this chapter I make a case for data curation. Sharing files implies a certain bulkiness of thought, and disembodies data from its meaning. Thus, I am looking at how changing the paradigm to an object-based, curatorial approach influences communication in the design process.

Read More: Chapter 5 Data Classification

Key Sections

5.1 Challenging Centralisation 5.2 From Sharing Files to Curating Data 5.4 Object Identity, Immutability & Data Deduplication

Does digital design data lend itself to multiple classifications?

Yes: on average, design models where Speckle was used were broken down into 2.47 separate sub-classifications (sources). Overall, the average count of both sources and receivers per model was 2.78.

Does the above result in a productive informational exchange?

Yes: on average, each single source has at least two receivers (ratio of 2.26 receivers per source). Information produced is consumed.

What are the implications on storage efficiency?

An object-based approach to data persistence is potentially twice as efficient as a file-based one in enabling multiple overlapping classifications of design information.

Data Transaction

Enabling Communicative Exchanges: Nextness, Sequentiality.

Message sequentiality (order) and "nextness" (closness) all contribute to a productive exchange of information - they are part of the communicative contract people engaged in dialogue enter into. In this chapter, I investigate how communicative contracts can be upheld—or shaped—in a digital environment.

Read More: Chapter 6 Data Transaction


6.1 Nextness & Sequentiality 6.2 Optimising Transactions: Differential Updates 6.3 Assembling the Network

To what extent can nextness in digital design communication be enabled?

Empirical observations show a five-fold improvement in transaction size. The upper performance limit is not bounded—as such, small transactions can be instantaneous, greatly improving the adjacency of data exchanges.

Does correlating sources and receivers enable a satisfactory level of sequentiality?

Yes, stakeholders, by seeing who is depending on their data, and whom they are dependent on, can successfully navigate design tasks and abstract a mental model of design dependencies.


I'm making three important points in this chapter. First, you can't have a productive design process without composability and ontological flexibility at the data level. Second, the single source of truth metaphor that plagues AEC is wrongly derived from a technical model of communication - it's time to let it go. Last, data centralisation won't work - it's against the DNA of AEC itself.

Read More: Chapter 7 Discussion

Composability trumps completeness

A single, centralised, complete object model is not required to have productive communication in digital design.

Composability should be prioritised over completeness/universality. AEC will never have, and does not need, one single standard.

7.1 Schemas and Standards

Single source of truth fallacy

The single source of truth design model, while valid for technical communication frameworks, is not a hard requirement in digital design communication.

Stakeholders can productively curate the information they need into leaner, federated models on a need by basis.

7.2 Classification and Curation

Data ownership & residency

Centralisation of data is in direct conflict with the distributed nature of the AEC industry. Data ownership is not a tangential concern: it's a key issue.

The reduced accessibility and closed source nature of data solutions for the built environment is holding AEC back.

7.4 Data Ownership

Contribution & Impact

I believe that the main contribution of my research project is an integrated technical and sociological rethink of communication in the digital design process that challenges the existing status quo of the AEC industry. Lastly, I show (both practically, through Speckle, and theoretically, through the analyses herein) that digital technologies can embrace the diversity and richness of the design process and open an accessible and ethical pathway towards a digitally integrated built environment.

Read More: Chapter 8 Conclusion

Speckle continues to grow and serve both practice and research.


This research has been conducted at the Bartlett School of Architecture, UCL, within the InnoChain Early Training Network supported by the European Union’s Horizon 2020 Marie Skłodowska-Curie grant agreement No. 642877.

This project would have not been possible without my supervisors, Sean Hanna, Bob Sheil and Robert Aish, whose guidance was invaluable. Similarly, I owe an huge debt of gratitude to Martin Tamke and Mette Ramsgaard-Thomsen who enabled the InnoChain project in the first place, as well as to the generosity of the persons that made this research network possible. Most importantly, to Luis Fraguada, who was a productive and calming presence at McNeel Europe, and to Giovanni Betti, the discussions with which could never end, at HENN Berlin.

Further thanks go to my mother (who showed me the way to Garfinkel, Grice, Sperber and Wilson), my father, my partner and my closest friends. Without your patience and support, this project would have not been done.

Dimitrie Stefanescu, London, October 2019