“Archaeology, Typology and Machine Epistemology” submitted by G. Lucas (1) offers a refreshing and welcome reflection on the role of computer-based practice, type-thinking and approaches to typology in the age of big data and the widely proclaimed ‘Third Science Revolution’ (2–4). At the annual meeting of the EAA in Maastricht in 2017, a special thematic block was dedicated to issues and opportunities linked to the Third Science Revolution in archaeology “because of [its] profound and wide ranging impact on practice and theory in archaeology for the years to come” (5). Even though the Third Science Revolution, as influentially outlined by Kristiansen in 2014 (2), has occasionally also been met with skepticism and critique as to its often implicit scientism and epistemological naivety (6–8), archaeology as a whole seems largely euphoric as to the promises of the advancing ‘revolution’. As Lucas perceptively points out, some even regard it as the long-awaited opportunity to finally fulfil the ambitions and goals of Anglophone processualism. The irony here, as Lucas rightly notes, is that early processualists initially foregrounded issues of theory and scientific epistemology, while much work conducted under the banner of the Third Science Revolution, especially within its computational branches, does not. Big data advocates have echoed Anderson’s much-cited “end of theory” (9) or at least emphatically called for an ‘empirization’ and ‘computationalization’ of theory, often under the banner of ‘data-driven archaeology’ (10), yet typically without much specification of what this is supposed to mean for archaeological theory and reflexivity. The latter is indeed often openly opposed by archaeological Third Science Revolution enthusiasts, arguably because it is viewed as part of the supposedly misguided ‘post-modernist’ project.

Lucas makes an original meta-archaeological contribution here and attempts to center the epistemological, ontological and praxeological dimensions of what is actually – in situated archaeological praxis and knowledge-production – put at stake by the mobilization of computers, algorithms and artificial intelligence (AI), including its many but presently under-reflected implications for ordering practices such as typologization. Importantly, his perspective thereby explicitly and deliberately breaks with the ‘normative project’ in traditional philosophy of science, which sought to nail down a universal, prescriptive way of doing science and securing scientific knowledge. He instead focuses on the practical dimensions and consequences of computer-reliant archaeologies, what actually happens on the ground as researchers try to grapple with the digital and the artefactual and try to negotiate new insights and knowledge, including all of the involved messiness – thereby taking up the powerful impetus of the broader practice turn in interdisciplinary science studies and STS (Science and Technology Studies (11)) (12–14), which have recently also re-oriented archaeological self-observation, metatheory and epistemology (15). This perspective on the dawning big data age in archaeology and incurred changes in the status, nature and aims of type-thinking produces a number of important insights, which Lucas fruitfully discusses in relation to promises of ‘automation’ and ‘novelty’ as these feature centrally in the rhetorics and politics of the Third Science Revolution. 

With regard to automation, Lucas makes the important point that machine or computer work as championed by big data proponents cannot adequately be qualified or understood if we approach the issue from a purely time-saving perspective. The question we have to ask instead is what work do machines actually do and how do they change the dynamics of archaeological knowledge production in the process? In this optic, automation and acceleration achieved through computation appear to make most sense in the realm of the uncontroversial, in terms of “reproducing an accepted way of doing things” as Lucas says, and this is precisely what can be observed in archaeological practice as well. The ramifications of this at first sight innocent realization are far-reaching, however. If we accept the noncontroversial claim that automation partially bypasses the need for specialists through the reproduction of already “pre-determined outputs”, automated typologization would primarily be useful in dealing with and synthesizing larger amounts of information by sorting artefacts into already accepted types rather than create novel types or typologies. If we identity the big data promise at least in part with automation, even the detection of novel patterns in any archaeological dataset used to construct new types cannot escape the fact that this novelty is always already prefigured in the data structure devised. The success of ‘supervised learning’ in AI-based approaches illustrates this. Automation thus simply shifts the epistemological burden back to data selection and preparation but this is rarely realized, precisely because of the tacit requirement of broad non-contentiousness. 

Minimally, therefore, big data approaches ironically curtail their potential for novelty by adhering to conventional data treatment and input formats, rarely problematizing the issue of data construction and the contested status of (observational) data themselves. By contrast, they seek to shield themselves against such attempts and tend to retain a tacit universalism as to the nature of archaeological data. Only in this way is it possible to claim that such data have the capacity to “speak for themselves”. To use a concept borrowed from complexity theory, archaeological automation-based type-construction that relies on supposedly basal, incontrovertible data inputs can only ever hope to achieve ‘weak emergence’ (16) – ‘strong emergence’ and therefore true, radical novelty require substantial re-thinking of archaeological data and how to construct them. This is not merely a technical question as sometimes argued by computational archaeologies – for example with reference to specifically developed, automated object tracing procedures – as even such procedures cannot escape the fundamental question of typology: which kind of observations to draw on in order to explore what aspects of artefactual variability (and why). The focus on readily measurable features – classically dimensions of artefactual form – principally evades the key problem of typology and ironically also reduces the complexity of artefactual realities these approaches assert to take seriously. The rise of computational approaches to typology therefore reintroduces the problem of universalism and, as it currently stands, reduces the complexity of observational data potentially relevant for type-construction in order to enable to exploration of the complexity of pattern. It has often been noted that this larger configuration promotes ‘data fetishism’ and because of this alienates practitioners from the archaeological record itself – to speak with Marxist theory that Lucas briefly touches upon. We will briefly return to the notion of ‘distance’ below because it can be described as a symptomatic research-logical trope (and even a goal) in this context of inquiry. 

In total, then, the aspiration for novelty is ultimately difficult to uphold if computational archaeologies refuse to engage in fundamental epistemological and reflexive self-engagement. As Lucas poignantly observes, the most promising locus for novelty is currently probably not to be found in the capacity of the machines or algorithms themselves, but in the modes of collaboration that become possible with archaeological practitioners and specialists (and possibly diverse other groups of knowledge stakeholders). In other words, computers, supercomputers and AI technologies do not revolutionize our knowledge because of their superior computational and pattern-detection capacities – or because of some mysterious ‘superintelligence’ – but because of the specific ‘division of labour’ they afford and the cognitive challenge(s) they pose. Working with computers and AI also often requires to ask new questions or at least to adapt the questions we ask. This can already be seen on the ground, when we pay attention to how machine epistemologies are effectively harnessed in archaeological practice (and is somewhat ironic given that the promise of computational archaeology is often identified with its potential to finally resolve "long-standing (old) questions"). The Third Science Revolution likely prompts a consequential transformation in the structural and material conditions of the kinds of ‘distributed’ processes of knowledge production that STS have documented as characteristic for scientific discoveries and knowledge negotiations more generally (14, 17, 18). This ongoing transformation is thus expected not only to promote new specializations with regard to the utilization of the respective computing infrastructures emerging within big data ecologies but equally to provoke increasing demand for new ways of conceptualizing observations and to reformulate the theoretical needs and goals of typology in archaeology. The rediscovery of reflexivity as an epistemic virtue within big data debates would be an important step into this direction, as it would support the shared goal of achieving true epistemic novelty, which, as Lucas points out, is usually not more than an elusive self-declaration. Big data infrastructures require novel modes of human-machine synergy, which simply cannot be developed or cultivated in an atheoretical and/or epistemological disinterested space. 

Lucas’ exploration ultimately prompts us to ask big questions (again), and this is why this is an important contribution. The elephant in the room, of course, is the overly strong notion of objectivity on which much computational archaeology is arguably premised – linked to the vow to eventually construct ‘objective typologies’. This proclivity, however, re-tables all the problematic debates of the 1960s and – to speak with the powerful root metaphor of the machine fueling much of causal-mechanistic science (19, 20) – is bound to what A. Wylie (21) and others have called the ‘view from nowhere’. Objectivity, in this latter view, is defined by the absence of positionality and subjectivity – chiefly human subjectivity – and the promise of the machine, and by extension of computational archaeology, is to purify and thus to enhance processes of knowledge production by minimizing human interference as much as possible. The distancing of the human from actual processes of data processing and inference is viewed as positive and sometimes even as an explicit goal of scientific development. Interestingly, alienation from the archaeological record is framed as an epistemic virtue here, not as a burden, because close connection with (or even worse, immersion in) the intricacies of artefacts and archaeological contexts supposedly aggravates the problem of bias. The machine, in this optic, is framed as the gatekeeper to an observer-independent reality – which to the backdoor often not only re-introduces Platonian/Aristotelian pledges to a quasi-eternal fabric of reality that only needs to be “discovered” by applying the right (broadly nonhuman) means, it is also largely inconsistent with defendable and currently debated conceptions of scientific objectivity that do not fall prey to dogma.  

Furthermore, current discussions on the open AI ChatGPT have exposed the enormous and still under-reflected dangers of leaning into radical renderings of machine epistemology: precisely because of the principles of automation and the irreducible theory-ladenness of all data, ecologies such as ChatGPT tend to reinforce the tacit epistemological background structures on which they operate and in this way can become collaborators in the legitimization and justification of the status quo (which again counteracts the potential for novelty) – they reproduce supposedly established patterns of thought. This is why, among other things, machines and AI can quickly become perpetuators of parochial and neocolonial projects – their supposed authority creates a sense of impartiality that shields against any possible critique. With Lucas, we can thus perhaps cautiously say that what is required in computational archaeology is to defuse the authority of the machine in favour of a new community archaeology that includes machines as (fallible) co-workers. Radically put, computers and AI should be recognized as subjects themselves, and treated as such, with interesting perspectives on team science and collaborative practice.

Bibliography

1. Lucas, G. (2022). Archaeology, Typology and Machine Epistemology. https:/doi.org/10.5281/zenodo.7620824.

2. Kristiansen, K. (2014). Towards a New Paradigm? The Third Science Revolution and its Possible Consequences in Archaeology. Current Swedish Archaeology 22, 11–34. https://doi.org/10.37718/CSA.2014.01.

3. Kristiansen, K. (2022). Archaeology and the Genetic Revolution in European Prehistory. Elements in the Archaeology of Europe. https://doi.org/10.1017/9781009228701

4. Eisenhower, M. S. (1964). The Third Scientific Revolution. Science News 85, 322/332. https://www.sciencenews.org/archive/third-scientific-revolution.

5. The ‘Third Science Revolution’ in Archaeology. http://www.eaa2017maastricht.nl/theme4 (March 16, 2023).

6. Ribeiro, A. (2019). Science, Data, and Case-Studies under the Third Science Revolution: Some Theoretical Considerations. Current Swedish Archaeology 27, 115–132. https://doi.org/10.37718/CSA.2019.06

7. Samida, S. (2019). “Archaeology in times of scientific omnipresence” in Archaeology, History and Biosciences: Interdisciplinary Perspectives, pp. 9–22. https://doi.org/10.1515/9783110616651

8. Sørensen, T. F.. (2017). The Two Cultures and a World Apart: Archaeology and Science at a New Crossroads. Norwegian Archaeological Review 50, 101–115. https://doi.org/10.1080/00293652.2017.1367031

9. Anderson, C. (2008). The end of theory: The data deluge makes the scientific method obsolete. Wired. https://www.wired.com/2008/06/pb-theory/.

10. Gattiglia, G. (2015). Think big about data: Archaeology and the Big Data challenge. Archäologische Informationen 38, 113–124. https://doi.org/10.11588/ai.2015.1.26155

11. Hackett, E. J. (2008). The handbook of science and technology studies, Third edition, MIT Press/Society for the Social Studies of Science.

12. Ankeny, R., Chang, H., Boumans, M. and Boon, M. (2011). Introduction: philosophy of science in practice. Euro Jnl Phil Sci 1, 303. https://doi.org/10.1007/s13194-011-0036-4

13. Soler, L., Zwart, S., Lynch, M., Israel-Jost, V. (2014). Science after the Practice Turn in the Philosophy, History, and Social Studies of Science, Routledge.

14. Latour, B. and Woolgar, S. (1986). Laboratory life: the construction of scientific facts, Princeton University Press.

15. Chapman, R. and Wylie, A. (2016) Evidential reasoning in archaeology, Bloomsbury Academic.

16. Greve, J. and Schnabel, A. (2011). Emergenz: zur Analyse und Erklärung komplexer Strukturen, Suhrkamp.

17. Shapin, S., Schaffer, S. and Hobbes, T. (1985). Leviathan and the air-pump: Hobbes, Boyle, and the experimental life, including a translation of Thomas Hobbes, Dialogus physicus de natura aeris by Simon Schaffer, Princeton University Press.

18. Galison, P. L. and Stump, D. J. (1996).The Disunity of Science: Boundaries, Contexts, and Power, Stanford University Press.

19. Pepper, S. C. (1972). World hypotheses: a study in evidence, 7. print, University of California Press.

20. Hussain, S. T. (2019). The French-Anglophone divide in lithic research: A plea for pluralism in Palaeolithic Archaeology, Open Access Leiden Dissertations. https://hdl.handle.net/1887/69812 

21. A. Wylie, A. (2015). “A plurality of pluralisms: Collaborative practice in archaeology” in Objectivity in Science, pp. 189-210, Springer. https://doi.org/10.1007/978-3-319-14349-1_10

The article submitted by Guillaume Porraz et al. [1] shed light on the evolutionary changes recorded during the Pre-Still Bay Lynn stratigraphic unit (SU) from Diepkloof (Southern Africa). It promotes a multi-proxy and integrative approach based on a set of innovative behaviors, such as the engraving of geometric forms, silcrete heat- treatment, the use of adhesive, bladelet and bifacial tools production. This approach is not so common in Middle Stone Age (MSA) studies and makes a lot of sense for discussing the mechanisms that have fostered later innovations during the Still Bay and Howiesons Poort periods. The various innovations that emerge synchronously in this layer contrast with earlier innovations which appear as isolated phenomena in the MSA archaeological record. The strong inventiveness documented in Lynn SU is reported to a phase of experimentation for testing new ideas, new behaviors that would have played a crucial role for the emergence of the Still Bay in a context of socio-economic transformation.

The data presented in this article broadens the scope of two previous articles [2-3] based on a more representative record, collected on an area of 3,5 m² opposed to 2 m² previously, and on the first presentation and description of an engraved bone with a rhomboid pattern. Macro- and microscopic analyses together with the analysis of the distribution of the engraved lines argue convincingly for an intentional engraving. This article constitutes a key contribution to the question of HOW emerged modern cultures in Southern Africa, while calling for further research related to sites’ function, environment and local resources to address the ever-debated question of WHY the MSA groups from Southern Africa developed such unprecedented inventiveness. It makes no doubt that this article deserves recommendation by PCI Archaeology.

[1] Porraz, G., Schmidt, P., Bereiziat, G., Brugal, J.Ph., Dayet, L., Igreja, M., Miller, C.E., Viola, C., Tribolo, C., Val, A., Verna, C., Texier, P.J. 2020. Experimentation preceding innovation in a MIS5 Pre-Still Bay layer from Diepkloof Rock Shelter (South Africa): emerging technologies and symbols. 10.32942/osf.io/ch53r

[2] Porraz, G., Texier, P.J., Archer, W., Piboule, M., Rigaud, J.P, Tribolo, C. 2013. Technological successions in the Middle Stone Age sequence of Diepkloof Rock Shelter, Western Cape, South Africa. Journal of Archaeological Science 40, 3376–3400. 10.1016/j.jas.2013.02.012

[3] Porraz, G., Texier, J.P. Miller, C.E., 2014. Le complexe bifacial Still Bay et ses modalités d’émergence à l’abri Diepkloof (Middle Stone Age, Afrique du Sud). In: XXVIIème Congrès Préhistorique de France, Transitions, Ruptures et Continuité en Préhistoire. Mémoires de la Société Préhistorique Française, 155–175.

The paper entitled “Comparing summed probability distributions of shoreline and radiocarbon dates from the Mesolithic Skagerrak coast of Norway” by Isak Roalkvam and Steinar Solheim (2024) sheds new light on the degree to which shoreline dating may be used as a reliable chronological and palaeodemographic proxy in the Mesolthic of southern Norway.

Based on geologically motivated investigations of eustatic and isostatic sea-level changes, shoreline dating has long been used as a method to date archaeological sites in Scandinavia, not least in Norway (e.g., Bjerck 2008; Astrup 2018). Establishing reliable sea-level curves requires much effort and variations across regions may be substantial. While this topic has seen a great deal of attention in Norway specifically, many purely geological questions remain. In addition, dating archaeological sites by linking their elevation to previously established seal-level curves relies strongly on the foundational assumption that such sites were in fact shore-bound. Given the strong contrast between terrestrial and marine productivity in high-latitude regions such as Norway, this assumption per se is not unreasonable. It is very likely that the sea has played a decisive role in the lives of Stone Age peoples throughout (Persson et al. 2017), just as it has in later periods here. However, many confounding factors relating to both taphonomy and human behaviour are also likely to have loosened the shore/site relationship. Systematic variations driven by cultural norms about settlement location, mobility, as well as factors such as shelter construction, fuel use and a range of other possible factors could variously have impacted the validity or at least the precision of shoreline dating.

By developing a new methodology for handling and assessing a large number of shoreline dated sites, Roalkvam and Solheim use state-of-the-art Bayesian statistical methods to compare shoreline and radiocarbon dates as proxies for population activity. The probabilistic treatment of shoreline dates in this way is novel, and the divergences between the two data sets are interpreted by the authors in light of specific behavioural, cultural, and demographic changes. Many of the peaks and troughs observed in these time-series may be interpreted in light of long-observed cultural transitions while others may relate to population dynamics now also visible in palaeogenomic analyses (Günther et al. 2018; Manninen et al. 2021). Overall, this paper makes an innovative and fresh contribution to the use of shoreline dating in Norwegian archaeology, specifically by articulating it with recent developments in Open Science and data-driven approaches to archaeological questions (Marwick et al. 2017).

References

Astrup, P. M. 2018. Sea-Level Change in Mesolithic Southern Scandinavia : Long- and Short-Term Effects on Society and the Environment. Aarhus: Aarhus University Press.

Bjerck, H. B. 2008. Norwegian Mesolithic Trends: A Review. In Mesolithic Europe, edited by Geoff Bailey and Penny Spikins, 60–106. Cambridge: Cambridge University Press.

Günther, T., Malmström, H., Svensson, E. M., Omrak, A., Sánchez-Quinto, F., Kılınç, G. M., Krzewińska, M. et al. 2018. Population Genomics of Mesolithic Scandinavia: Investigating Early Postglacial Migration Routes and High-Latitude Adaptation. PLOS Biology 16 (1): e2003703. https://doi.org/10.1371/journal.pbio.2003703

Manninen, M. A., Damlien, H., Kleppe, J. I., Knutsson, K., Murashkin, A., Niemi, A. R., Rosenvinge, C. S. and Persson, P. 2021. First Encounters in the North: Cultural Diversity and Gene Flow in Early Mesolithic Scandinavia. Antiquity 95 (380): 310–28. https://doi.org/10.15184/aqy.2020.252

Marwick, B., d’Alpoim Guedes, J. A., Barton, C. M., Bates, L. A., Baxter, M., Bevan, A., Bollwerk, E. A. et al. 2017. Open Science in Archaeology. The SAA Archaeological Record 17 (4): 8–14. https://doi.org/10.31235/osf.io/72n8g

Persson, P., Riede, F., Skar, B., Breivik, H. M. and Jonsson, L. 2017. The Ecology of Early Settlement in Northern Europe: Conditions for Subsistence and Survival. Sheffield: Equinox.

Roalkvam, I. and Solheim, S. (2024). Comparing summed probability distributions of shoreline and radiocarbon dates from the Mesolithic Skagerrak coast of Norway, SocArXiv, 2f8ph, ver. 5 peer-reviewed and recommended by Peer Community in Archaeology. https://doi.org/10.31235/osf.io/2f8ph

Studies of the economy of the Roman Empire have become increasingly interdisciplinary and nuanced in recent years, allowing the discipline to make great strides in data collection and importantly in the methods through which this increasing volume of data can be effectively and meaningfully analysed [see for example 1 and 2]. One of the key aspects of modelling the ancient economy is understanding movement and transport costs, and how these facilitated trade, communication and economic development. With archaeologists adopting more computational techniques and utilising GIS analysis beyond simply creating maps for simple visualisation, understanding and modelling the costs of traversing archaeological landscapes has become a much more fruitful avenue of research. Classical archaeologists are often slower to adopt these new computational techniques than others in the discipline. This is despite (or perhaps due to) the huge wealth of data available and the long period of time over which the Roman economy developed, thrived and evolved. This all means that the Roman Empire is a particularly useful proving ground for testing and perfecting new methodological developments, as well as being a particularly informative period of study for understanding ancient human behaviour more broadly. This paper by Page [3] then, is well placed and part of a much needed and growing trend of Roman archaeologists adopting these computational approaches in their research. 

Page’s methodology builds upon De Soto’s earlier modelling of transport costs [4] and applies it in a new setting. This reflects an important practice which should be more widely adopted in archaeology. That of using existing, well documented methodologies in new contexts to offer wider comparisons. This allows existing methodologies to be perfected and tested more robustly without reinventing the wheel. Page does all this well, and not only builds upon De Soto’s work, but does so using a case study that is particularly interesting with convincing and significant results. 

As Page highlights, Northern Italy is often thought of as relatively isolated in terms of economic exchange and transport, largely due to the distance from the sea and the barriers posed by the Alps and Apennines. However, in analysing this region, and not taking such presumptions for granted, Page quite convincingly shows that the waterways of the region played an important role in bringing down the cost of transport and allowed the region to be far more interconnected with the wider Roman world than previous studies have assumed.  

This article is clearly a valuable and important contribution to our understanding of computational methods in archaeology as well as the economy and transport network of the Roman Empire. The article utilises innovative techniques to model transport in an area of the Roman Empire that is often overlooked, with the economic isolation of the area taken for granted. Having high quality research such as this specifically analysing the region using the most current methodologies is of great importance. Furthermore, developing and improving methodologies like this allow for different regions and case studies to be analysed and directly compared, in a way that more traditional analyses simply cannot do. As such, Page has demonstrated the importance of reanalysing traditional assumptions using the new data and analyses now available to archaeologists. 

References

[1] Brughmans, T. and Wilson, A. (eds.) (2022). Simulating Roman Economies: Theories, Methods, and Computational Models. Oxford. 

[2] Dodd, E.K. and Van Limbergen, D. (eds.) (2024). Methods in Ancient Wine Archaeology: Scientific Approaches in Roman Contexts. London ; New York. 

[3] Page, J. (2024). Rivers vs. Roads? A route network model of transport infrastructure in Northern Italy during the Roman period, Zenodo, 7971399, ver. 3 peer-reviewed and recommended by Peer Community in Archaeology. https://doi.org/10.5281/zenodo.7971399

[4] De Soto P (2019). Network Analysis to Model and Analyse Roman Transport and Mobility. In: Finding the Limits of the Limes. Modelling Demography, Economy and Transport on the Edge of the Roman Empire. Ed. by Verhagen P, Joyce J, and Groenhuijzen M. Springer Open Access, pp. 271–90. https://doi.org/10.1007/978-3-030-04576-0_13

The paper by Brown & Lewis [1] presents an approach to measure seasonal mobility and subsistence practices. In order to do so, the paper proposes a Bayesian mixture model to estimate the annual distribution of shellfish harvesting activity. Following the recommendations of the two reviewers, the paper presents a clear and innovative method to assess seasonal mobility for prehistoric groups, although it could benefit from additional references regarding isotopic literature.

While the adequacy of isotope analysis for estimating mobility patterns in Archaeology has been extensively proven by now, work on specific seasonal mobility is not that much abundant. However, this is a key issue, since seasonal mobility is one of the main social components defining the differences between groups both considering farming vs hunting and gathering or even among hunter-gatherer groups themselves. In this regard, the paper brings a valuable methodological resources that can be used for further research in this issue.

One of its greatest values is the fact that it can quantify the uncertainty present in previous isotope studies in seasonal mobility. As stated by the authors, the model can still undergo several optimisation aspects, but as it stands, it is already providing a valuable asset regarding the quantification of uncertainy in the isotopic studies of seasonal mobility.

Reference

[1] Brown, J. and Lewis, G. (2024). Inferring shellfishing seasonality from the isotopic composition of biogenic carbonate: A Bayesian approach. Zenodo, 7949547, ver. 3 peer-reviewed and recommended by Peer Community in Archaeology. https://doi.org/10.5281/zenodo.7949547