As part of my PhD project, over the last few months I have been researching trends in Visualisation from its incipit to today. I’m attempting to create a solid theoretical background upon which to build my own work, and in the process I am learning a lot of the philosophical and methodological foundations upon which 3D reconstruction stands.
There are many topics that I have found particularly interesting, some of which I have talked about before, and some of which I will write about in the near future. Today though I would like to concentrate on a topic I feel has only been discussed in passing: the creation of “alternative models”. Not only do I feel we need to dwell upon this subject more, I want to go against current trend to present a counterargument for their utility.
3D researchers have been advocating for alternative models from the start of Visualisation (Reilly 1992; Mathur 1997; Roberts and Ryan 1997; The Guardian 1999; Huggett and Guo-Yuan 2000). The basic idea is that it is impossible to present all hypotheses in a single 3D reconstruction, so a number of models should be reconstructed instead, each representing an alternative theory.
Alternative models fit within the general concept of accuracy. One of the major concerns with 3D reconstruction is the narrow scope for presentation of hypothetical data. Models are often built upon incomplete information, and as a result it is impossible to recreate a perfectly accurate representation of the past. There is a distinct worry amongst specialists and skeptics alike that without sufficient transparency 3D reconstructions may misinform and deceive an uninformed user (Bayliss 2003; Kensek et al. 2004; amongst many others). This concern is certainly founded, and while some suggestions have been put forward in an attempt to minimise the problem (Pang et al. 1997; Strothotte et al. 1999), the lack of a cohesive and enforced set of principles limits the reliability of 3D reconstruction as a methodology. We are moving in a positive direction, and publications such as the London Charter offer legitimisation of the use of these technologies (Beacham et al. 2006; Denard 2009). Yet there is still a need for enforced guidelines that can reduce, or at least explicitly state. inaccuracies.
The use of pink cement (Lock 2003), or Dell’Unto et al.’s (2013) levels of accuracy are good approaches to the problem, and the literature has embraced such ideas for the better. Alternative models, on the other hand, have been lurking in the background, often mentioned but never fully discussed. They are mostly used as an addendum or a failsafe, an attempt to silence any possible critic of the accuracy of the models. It is interesting how most of the publications that mention alternative models do not present alternative models themselves: papers concerned with the theoretical background use it as an example of ways to preserve accuracy, while technical papers omit alternative models altogether (with a few exceptions i.e. Roussou and Drettakis 2003). It is also important to note that they are always mentioned positively.
I would like to present a counterargument. I do not believe alternative models are bad, or that they have no uses. There are certainly occasions in which they can convey information more efficiently than other methods. I do however believe that they have a limited scope, and that using them as a way to present inaccuracies is counterproductive to the defining of a 3D methodology.
The limitations of alternative models can be expressed as so:
- Physical and publication space managment
- Time requirements
- Multiple hypotheses representation
Physical and publication space management: using alternative models to present inaccuracies to the public or to fellow researchers requires for this information to be readily accessible. At present the issue with 3D documentation is one of space. Publications often do not possess enough space for a full documentation of the reconstruction process, and the problem is exacerbated in heritage management where information has to be tailored to an uninformed public. In most cases, the presentation needs cannot accommodate the presence of multiple models. Metadata and paradata are beginning to appear in publications through the use of online repositories, but the handling of multiple large models is still problematic.
Time requirements: 3D reconstruction is a process, which follows a number of steps. Work by Guidi et al. (2012; 2014) show an ideal example of the reconstruction procedure, with accumulative levels of detail and archaeological checks at the end of each phase. In order to reconstruct most alternative models, the split must occur right from the volumetric model. By doing so each variant must be constructed individually, or the subsequent changes must be reflected in each of the different models. Attempting to change the model in the later stages of production is still possible, but it requires more work, as there are more elements that need to be manipulated. Either way, the results is a distinct increase in the production time.
Multiple hypotheses representation: another logistical problem has to do with models with many conjectures. When is it necessary to create multiple models? If we are too strict with our definition of inaccuracies then every element in the reconstruction is in question. We would therefore end with numerous models with very subtle differences. On the other hand, if we we were too lax the purpose of using alternative models is void, as displaying only some of the hypotheses would render this process redundant. Additionally, if many hypotheses were to be displayed through alternative models, it would be necessary to create a model with every possible arrangement of hypotheses, exponentially increasing the number of reconstructions.
In addition to these limitations, my argument against the misuse of alternative models has to do with traditional methods of presenting archaeological data. Archaeological reports often attempt an objective analysis of the archaeological evidence, describing what has been found and trying to limit conjectures, frequently presenting multiple theories. Yet when interpreting an archaeological site, it is not uncommon to create a narrative. Certain evidence is omitted and new meaning is imposed on the remains, in an attempt to justify the current archaeological interpretation. Part of the scientific method is to present the data and propose a unifying theory, with the expectation that the theory shall change when new evidence comes to light. And archaeology subscribes to this view: while multiple scenarios may be presented in publication, there is often one preferred interpretation.
Therefore, in traditional archaeology the data is analysed to produce the results. Objective data is transformed into subjective interpretation, and with a change in the data comes a change in the interpretation. The uncertainty is accounted for in the source and not the output and the same should be true for 3D reconstruction. Metadata and paradata are paramount for the replicability of the process and for validating the hypotheses shown in the results. We must find better ways of preserving and presenting this form of information, to ensure the 3D methodology is valid. Alternative models however are part of the output, where subjectivity is allowed and even encouraged.
Critics often note that in Heritage Management the public is more susceptible to misinformation, making 3D reconstruction a ‘dangerous’ tool. Yet museum displays often present a single narrative, reducing complex archaeological issues to simplistic linear stories. While it is still a problem 3D reconstruction needs to address, the discussion has wider implications in the way we present data to the public.
Additionally, I am not saying that alternative models cannot have uses in archaeological presentation. In some cases the conflicting theories are at the core of the discussion, and alternative models help create a virtual representation of what this conflict appears like visually. A good example of this is the Patay-Horvath (2014) paper regarding the positioning of statues at the Temple of Zeus at Olympia. Here the alternative models are used effectively to communicate the main argument. It was however a deliberate choice from the author, and it had a specific purpose.
In conclusion, alternative models are not always the best option for presenting inaccuracies. While occasionally they can be effective in demonstrating conflicting hypotheses, in a wider methodology for 3D reconstruction they have little space. Presenting metadata and paradata still offers the best course for demonstrating uncertainty.