INTRODUCTION

The UNESCO World Heritage Convention (WHC) is a landmark international agreement aimed at identifying and preserving natural and cultural sites with outstanding universal value. First adopted in 1972, the WHC has been ratified by 195 state parties. As of July 2024, the World Heritage List (WHL) included 1223 inscribed sites, representing “an extraordinary combination of the unique and the representative” of the world’s heritage (Titchen 1996:240). A remarkable achievement of the WHC is recognizing that natural and cultural heritage have equal standing (Batisse and Bolla 2005). However, the selection criteria for including sites on the list were established based on a deeply ingrained distinction between nature and culture that has yet to be reconciled (Batisse and Bolla 2005, Larsen and Wijesuriya 2015). To bridge the nature-culture divide, solutions have been proposed with rather unsatisfactory results. Initially, the WHC introduced the concept of designating sites as “mixed cultural and natural heritage” or simply “mixed” (UNESCO 2023). Sites in this category must meet at least one criterion from both the cultural (i-vi) and natural (vii-x) categories (https://whc.unesco.org/en/criteria/). Currently, only 40 sites (3% of all World Heritage Sites) are classified as mixed, reflecting the limitations of this categorization (Scazzosi 2002).

In response to early dissatisfaction with the mixed category, which Fowler (2003:17) later described as an “intellectually flaccid idea,” the WHC introduced the cultural landscapes concept in 1992. This concept aims to recognize sites with significant interactions between people and the environment (Rössler 2006), described in the WHC as the “combined works of nature and man” (UNESCO 2023:22). However, cultural landscapes are rooted within dated Eurocentric paradigms of man over nature, as exemplified by Carl Sauer’s 20th-century quote: “Culture is the agent, the natural area is the medium, the cultural landscape the result” (Sauer 1925:46). This perspective contrasts with more holistic relationships to the land, such as those held by many indigenous peoples, where intangible cultural values are inextricably linked to the natural world and often lack the material expressions emphasized under the WHC (Fogarty 2020). Thus, although cultural landscapes can be designated under either the cultural or mixed categories, the criteria primarily emphasize material cultural aspects such as traditional human settlements, making it operationally more convenient to classify cultural landscapes within the cultural category (Aplin 2007). As a result, of 129 cultural landscapes (11% of all World Heritage Sites), only 10 are classified as mixed, when arguably there should be a much larger overlap (Fowler 2003).

To address the limitations of the mixed and cultural landscape concepts, adopting a biocultural perspective offers a promising way forward (Hanspach et al. 2020, Lukawiecki et al. 2022). The Florence Declaration (ICOMOS 2014) marked a significant advancement on this front, by recognizing that the interaction between culture and nature primarily occurs at the landscape level. This led to the concept of biocultural landscapes (Agnoletti and Rotherham 2015) because “most, if not all, landscapes are blends of human activity with the expression of biodiversity — that is, they are biocultural landscapes.” (Bridgewater and Walton 1996, as cited in Bridgewater and Rotherham 2019). Furthermore, it has been recognized that biocultural heritage emerges from this outcome of human-environment interactions, manifesting not only in tangible landscape features, but also in memories and living practices (Lindholm and Ekblom 2019). This underscores the necessity to adopt a holistic approach to heritage recognition that can effectively bridge the gap between nature and culture. Our study addresses this need by evaluating the explicit recognition of biocultural heritage through a comprehensive recategorization of all the World Heritage Sites.

METHODS

We aim to reclassify the World Heritage Sites using a recent definition of biocultural heritage proposed by Lindholm and Ekblom (2019): “Biocultural heritage comprises local ecological knowledge and practices, and associated ecosystems and biological resources (from genetic variation and species biodiversity), to the formation of landscape features and cultural landscapes, as well as the heritage, memory and living practices of the humanly built or managed environments.”

To conduct our analyses, we sourced official site descriptions from the WHL website (https://whc.unesco.org/en/list/) and evaluated whether site descriptions adhere to the Lindholm and Ekblom definition. For this, we employed large language models (LLMs), state-of-the-art artificial intelligence models that operate using deep learning with neural networks to understand, analyze, and generate text with advanced capabilities. Consequently, LLMs have demonstrated outstanding performance across a diverse range of natural language processing (NLP) tasks (Gilardi et al. 2023, Peña et al. 2023, Chang et al. 2024, Kojima et al. 2024, Törnberg 2024).

We employed a zero-shot classification approach utilizing LLMs to assess site descriptions without the need for any previously classified examples. This method capitalizes on the semantic reasoning capabilities of LLMs (Kojima et al. 2024), which can outperform fine-tuned models and even human classifiers—the current standard in text classification (Gilardi et al. 2023, Törnberg 2024). To assess the reliability of LLMs in identifying biocultural sites, we compared classifications from three leading LLMs (as of February 2024) and both authors (RDP and SG) on a random sample of 120 sites (Appendix 1). This comparison allowed us to verify consistency in the number of identified biocultural sites and select the LLM most closely aligned with human assessments for comprehensive reclassification of all World Heritage Sites. The three LLMs used were OpenAI’s ChatGPT-4, Anthropic’s Claude-2.1, and Google Gemini’s Ultra 1.0. Although LLMs are continually updated, the assigned classification task is straightforward for their current capabilities, and we anticipate that the results will hold under future advancements.

We tested how LLMs responded to different prompts and chose to operationalize the definition of biocultural heritage using two main criteria, each with two sub-criteria. This reduced ambiguity by having the LLMs assess each sub-criterion separately, leading to more consistent classifications. A site was classified as biocultural if it satisfied at least one sub-criterion from the first criteria (local ecological knowledge/practices or associated ecosystems/biological resources) and at least one of the second criteria (formation of landscape features/cultural landscapes or heritage/memory of human-managed environments). To ensure these classifications were firmly grounded in the provided content, the prompts instructed the LLMs to extract key phrases from site descriptions (Box 1). We then evaluated inter-rater agreement using Cohen’s kappa (κ) and the percentage agreement between the LLMs and a consensus classification established by both authors.

Based on our initial findings (see results), we selected Claude-2 to reclassify all World Heritage Sites (1199 listed as of February 2024). Site descriptions were formatted using XML tags to enable efficient processing, leveraging Claude-2’s ability to parse structured information. We further optimized the prompt for Claude-2 by restructuring the biocultural heritage definition into the two main criteria without sub-criteria (Box 2). To evaluate the consistency of Claude-2 classifications, we conducted three independent runs (Appendix 2) and measured intra-rater agreement using Fleiss’ Kappa (Fleiss et al. 2003). We adopted a majority-vote approach to determine biocultural status, classifying a site as biocultural if it was identified as such in at least two out of three runs. This approach was preferred over unanimous agreement to avoid potential exclusions due to subtle differences in interpretation. Additionally, 24 sites added to the WHL in July 2024 were reclassified by the authors, bringing the total to 1223 sites (Fig. 1).

RESULTS

Our results demonstrated strong consistency between the three LLMs and the two authors in identifying the potential number of biocultural sites in the WHL. In a random sample of 120 sites, the percentage of sites classified as biocultural were: RDP (32%), SG (28%), Claude-2 (29%), Gemini Ultra (34%), and ChatGPT-4 (51%). The substantially higher percentage from ChatGPT-4 indicated a potential susceptibility to acquiescence bias or “yes-saying” (Tjuatja et al. 2024), and we excluded its use from further analyses. A human consensus between RDP and SG identified 30% of the sites as biocultural. There was a good inter-rater agreement among Claude-2, Gemini Ultra, and the human consensus (κ = 0.57, 66.7% agreement). Claude-2 showed higher agreement with the human consensus (κ = 0.58, 82.5%) than Gemini Ultra (κ = 0.52, 79.2%) and was selected for reclassifying World Heritage Sites. Both LLMs showed stronger agreement with the human consensus for sites not classified as biocultural (Claude-2: 77.9%, Gemini Ultra: 73.4%) than for sites classified as biocultural (Claude-2: 54.3% agreement, Gemini Ultra: 50.9% agreement).

The reclassification of all World Heritage Sites using Claude-2 demonstrated excellent intra-rater agreement across three runs (κ = 0.7, 83.9% agreement). When a majority-vote approach was applied, 251 of the 1199 sites evaluated as of February 2024 were considered biocultural (21%). This included 79% of mixed sites and 60% of cultural landscapes. Building on the premise that all mixed sites and cultural landscapes can be considered biocultural and incorporating seven of the 24 additional sites added in July 2024 (reclassified by the authors), we estimate that up to 314 sites—or 26% of all World Heritage Sites—fall into the biocultural category (Fig. 1, Table 1, Appendix 3). This estimate closely aligns with results from the random sample of 120 sites assessed by humans and LLMs.

DISCUSSION

In this study, we propose a biocultural reclassification of the World Heritage Sites. Our findings reveal that up to 26% of sites could be classified as biocultural, leveraging LLMs to evaluate whether site descriptions in the WHL adhere with a recent definition of biocultural heritage (Lindholm and Ekblom 2019). Importantly, comparing the classifications from three LLMs and both authors revealed similar proportions of identified biocultural sites. This underscores the robustness of the biocultural heritage concept and its potential to reclassify World Heritage Sites, many of which hold unrecognized biocultural value. However, we found significant differences in the classification of individual sites, highlighting the need to refine the criteria for improved agreement. Thus, we advocate for the development and implementation of a comprehensive biocultural reclassification framework, which could advance UNESCO’s Global Strategy for a Representative, Balanced, and Credible World Heritage List, which has not been achieved so far (Steiner and Frey 2012, Ishizawa and Westrik 2021).

A biocultural reclassification of the WHL would offer the advantage of preserving the distinction between strictly cultural and natural sites as originally conceived, while introducing a biocultural category that effectively bridges the gap between them, explicitly recognizing the interconnectedness of nature and culture. A biocultural category would enable a more inclusive and diverse representation of global heritage (Table 1) because it can effectively represent the varying ways in which heritage is conceptualized, valued, and represented across different regions of the world. For instance, it can address the differences in heritage conceptualizations between Asia and Europe (Hutcheson 2021), the lack of inclusion of African heritage (Labadi 2022), and the disproportionate number of sites in Europe and North America on the WHL (Table 1). The biocultural category could also help overcome long-held landscape paradigms perpetuated by European dominance, such as intentionally designed landscapes like Baroque parks and gardens surrounding palaces or the preponderance of vineyards over other agricultural landscapes (Brumann and Gfeller 2022). The biocultural perspective is also essential for understanding the future of heritage (Holtorf and Högberg 2021) as we navigate the Anthropocene and the significant human impact on Earth’s systems (IPBES 2019).

The inclusion of a biocultural category is not without challenges. A key hurdle is the need to revise the conditions of integrity and authenticity, at least one of which must be met, for recognition and inscription on the WHL (UNESCO 2023). These conditions are important guiding frameworks for both the cultural and natural criteria, but applying them consistently has proven difficult (Alberts and Hazen 2010), even more so given the challenges of their application under current human impacts (Khalaf 2021, Fuhrmann 2022). A biocultural framework offers the potential to reconceptualize both integrity and authenticity by considering them as emergent properties of long-term dynamic interactions between humans and the natural environment, rather than from cultural or natural attributes in isolation (Gullino and Larcher 2013). Crucially, under this conceptualization indigenous and local communities are paramount in shaping and safeguarding this heritage (Swiderska et al. 2009), which might be threatened by cultural severance, the disruption of traditional practices and knowledge (Rotherham 2015).

The recognition of a biocultural category will therefore contribute toward an “Increased consideration of sustainable development through connecting conservation to communities” (UNESCO 2011:7) and align the World Heritage Convention with the global goals of the Kunming-Montreal Global Biodiversity Framework (https://www.cbd.int/gbf) through biocultural approaches to conservation (Gavin et al. 2015, Caillon et al. 2017). A strengthened biocultural perspective within UNESCO could also foster a closer integration between World Heritage Sites and Biosphere Reserves. The latter, administered by the Man and the Biosphere (MAB) program, are defined as “learning places for sustainable development” (https://www.unesco.org/en/mab/wnbr/about) with outstanding biocultural value (Bridgewater 2002, Cusens et al. 2022, Eliasson et al. 2023). This integration would be complementary and promising (Engelbauer 2023), reinforcing the MAB program’s goal to develop a scientific foundation for improving the relationship between humans and their environments. Ultimately, this will also enhance the capacity to address governance and management challenges more effectively (Hutcheson 2021).

In conclusion, our work supports recent calls to recognize biocultural heritage under the World Heritage Convention (Bhati and Epstein 2023). This recognition could bring up to a quarter of World Heritage Sites as biocultural and will strengthen the identification, management, and preservation of global heritage with outstanding universal value. Furthermore, the biocultural perspective centers the role of local and indigenous communities in shaping and maintaining these values. It also fosters more inclusive and community-centered approaches to heritage management and conservation strategies, integrating cultural practices with environmental protection and ensuring a holistic preservation of heritage essential for sustainable development amidst rapid societal and environmental changes.

LITERATURE CITED

Agnoletti, M., and I. D. Rotherham. 2015. Landscape and biocultural diversity. Biodiversity and Conservation 24(13):3155-3165. https://doi.org/10.1007/s10531-015-1003-8

Alberts, H. C., and H. D. Hazen. 2010. Maintaining authenticity and integrity at Cultural World Heritage Sites. Geographical Review 100(1):56-73. https://doi.org/10.1111/j.1931-0846.2010.00006.x

Aplin, G. 2007. World Heritage cultural landscapes. International Journal of Heritage Studies 13(6):427-446. https://doi.org/10.1080/13527250701570515

Batisse, M., and G. Bolla. 2005. The invention of “world heritage.” Translated by Robert Grauman and Johnson Raymond. History Papers, Association of Former UNESCO Staff Members, Paris, France.

Bhati, H. V., and Y. Epstein. 2023. Protection of biocultural heritage in the Anthropocene: towards reconciling natural, cultural, tangible and intangible heritage. Journal of Environmental Law 35(3):353-375. https://doi.org/10.1093/jel/eqad020

Bridgewater, P. B. 2002. Biosphere reserves: special places for people and nature. Environmental Science & Policy 5(1):9-12. https://doi.org/10.1016/S1462-9011(02)00018-7

Bridgewater, P., and I. D. Rotherham. 2019. A critical perspective on the concept of biocultural diversity and its emerging role in nature and heritage conservation. People and Nature 1(3):291-304. https://doi.org/10.1002/pan3.10040

Brumann, C., and A. É. Gfeller. 2022. Cultural landscapes and the UNESCO World Heritage List: perpetuating European dominance. International Journal of Heritage Studies 28(2):147-162. https://doi.org/10.1080/13527258.2021.1941197

Caillon, S., G. Cullman, B. Verschuuren, and E. J. Sterling. 2017. Moving beyond the human-nature dichotomy through biocultural approaches: Including ecological well-being in resilience indicators. Ecology and Society 22(4):27. https://doi.org/10.5751/ES-09746-220427

Chang, Y., X. Wang, J. Wang, Y. Wu, L. Yang, K. Zhu, H. Chen, X. Yi, C. Wang, Y. Wang, W. Ye, Y. Zhang, Y. Chang, P. S. Yu, Q. Yang, and X. Xie. 2024. A survey on evaluation of large language models. ACM Transactions on Intelligent Systems and Technology 15(3):39:1-45. https://doi.org/10.1145/3641289

Cusens, J., A. M. D. Barraclough, and I. E. Måren. 2022. Participatory mapping reveals biocultural and nature values in the shared landscape of a Nordic UNESCO Biosphere Reserve. People and Nature 4(2):365-381. https://doi.org/10.1002/pan3.10287

Eliasson, I., S. Fredholm, I. Knez, E. Gustavsson, and J. Weller. 2023. Cultural values of landscapes in the practical work of biosphere reserves. Land 12(3):587. https://doi.org/10.3390/land12030587

Engelbauer, M. 2023. Global assessment of recent UNESCO Biosphere Reserve quality enhancement strategies and interlinkages with other UNESCO labels. Würzburg University Press, Würzburg, Germany.

Fleiss, J. L., B. Levin, and M. C. Paik. 2003. Statistical methods for rates and proportions. Third edition. John Wiley & Sons, Hoboken, New Jersey, USA. https://doi.org/10.1002/0471445428

Fogarty, I. 2020. Untangling the concept of cultural landscapes: a critical review. Pages 17-30 in A. de Carvalho Antunes, G. Angjeliu, and M. Bellanova, editors. Advances in cultural heritage studies, year 2020. Contributions of the European Students’ Association for Cultural Heritage. Mazu Press, Oeiras, Portugal.

Fowler, P. J. 2003. World heritage cultural landscapes 1992-2002. World Heritage Papers 6. Cultural landscapes: The challenges of conservation. UNESCO, Paris, France.

Fuhrmann, C. 2022. Sustainability - a guiding principle of the world heritage convention - What has been achieved - what is missing - what is the future perspective. Pages 445-457 in M.-T. Albert, R. Bernecker, C. Cave, A. C. Prodan, and M. Ripp, editors. 50 Years World Heritage Convention: shared responsibility - conflict & reconciliation. Springer International Publishing, Cham, Switzerland. https://doi.org/10.1007/978-3-031-05660-4_35

Gavin, M. C., J. McCarter, A. Mead, F. Berkes, J. R. Stepp, D. Peterson, and R. Tang. 2015. Defining biocultural approaches to conservation. Trends in Ecology & Evolution 30(3):140-145. https://doi.org/10.1016/j.tree.2014.12.005

Gilardi, F., M. Alizadeh, and M. Kubli. 2023. ChatGPT outperforms crowd workers for text-annotation tasks. Proceedings of the National Academy of Sciences 120(30):e2305016120. https://doi.org/10.1073/pnas.2305016120

Gullino, P., and F. Larcher. 2013. Integrity in UNESCO World Heritage Sites. A comparative study for rural landscapes. Journal of Cultural Heritage 14(5):389-395. https://doi.org/10.1016/j.culher.2012.10.005

Hanspach, J., L. Jamila Haider, E. Oteros-Rozas, A. Stahl Olafsson, N. M. Gulsrud, C. M. Raymond, M. Torralba, B. Martín-López, C. Bieling, M. García-Martín, C. Albert, T. H. Beery, N. Fagerholm, I. Díaz-Reviriego, A. Drews-Shambroom, and T. Plieninger. 2020. Biocultural approaches to sustainability: a systematic review of the scientific literature. People and Nature 2(3):643-659. https://doi.org/10.1002/pan3.10120

Holtorf, C., and A. Högberg, editors. 2021. Cultural heritage and the future. Routledge, London, UK. https://doi.org/10.4324/9781315644615

Hutcheson, A. 2021. Divergent heritages? UNESCO and the cultural heritage of landscapes in the UK and Japan. Historic Environment: Policy & Practice 12(2):164-185. https://doi.org/10.1080/17567505.2021.1903147

Intergovernmental Science-Policy Platform on Biodiversity and Ecosystem Service (IPBES). 2019. Summary for policymakers of the global assessment report on biodiversity and ecosystem services of the Intergovernmental Science-Policy Platform on Biodiversity and Ecosystem Service. IPBES secretariat, Bonn, Germany.

International Council on Monuments and Sites (ICOMOS). 2014. The Florence Declaration on Heritage and Landscape as Human Values. ICOMOS, Charenton-le-Pont, France

Ishizawa, M., and C. Westrik. 2021. Analysis of the global strategy for a representative, balanced and credible World Heritage List (1994-2020). World Heritage Centre, Paris, France.

Khalaf, R. W. 2021. World Heritage on the move: abandoning the assessment of authenticity to meet the challenges of the twenty-first century. Heritage 4(1):371-386. https://doi.org/10.3390/heritage4010023

Kojima, T., S. S. Gu, M. Reid, Y. Matsuo, and Y. Iwasawa. 2024. Large language models are zero-shot reasoners. Pages 22199-22213 in Proceedings of the 36th International Conference on Neural Information Processing Systems. Curran Associates Inc., Red Hook, New York, USA.

Labadi, S. 2022. The World Heritage Convention at 50: management, credibility and sustainable development. Journal of Cultural Heritage Management and Sustainable Development 14(4):750-764. https://doi.org/10.1108/JCHMSD-05-2022-0077

Larsen, P., and G. Wijesuriya. 2015. Nature-culture interlinkages in World Heritage: bridging the gap. World Heritage 75:4-15.

Lindholm, K.-J., and A. Ekblom. 2019. A framework for exploring and managing biocultural heritage. Anthropocene 25:100195. https://doi.org/10.1016/j.ancene.2019.100195

Lukawiecki, J., J. Wall, R. Young, J. Gonet, G. Azhdari, and F. Moola. 2022. Operationalizing the biocultural perspective in conservation practice: a systematic review of the literature. Environmental Science & Policy 136:369-376. https://doi.org/10.1016/j.envsci.2022.06.016

Peña, A., A. Morales, J. Fierrez, I. Serna, J. Ortega-Garcia, Í. Puente, J. Córdova, and G. Córdova. 2023. Leveraging large language models for topic classification in the domain of public affairs. Pages 20-33 in M. Coustaty and A. Fornés, editors. Document analysis and recognition - ICDAR 2023 Workshops. Springer Nature, Cham, Switzerland. https://doi.org/10.1007/978-3-031-41498-5_2

Rössler, M. 2006. World Heritage cultural landscapes: a UNESCO flagship programme 1992 - 2006. Landscape Research 31(4):333-353. https://doi.org/10.1080/01426390601004210

Rotherham, I. D. 2015. Bio-cultural heritage and biodiversity: emerging paradigms in conservation and planning. Biodiversity and Conservation 24(13):3405-3429. https://doi.org/10.1007/s10531-015-1006-5

Sauer, C. O. 1925. The morphology of landscape. University of California Publications in Geography 2(2):19-53.

Scazzosi, L. 2002. Landscape and cultural landscape: European landscape convention and UNESCO Policy. Cultural landscapes: The challenges of conservation. World Heritage Papers 7:55.

Steiner, L., and B. S. Frey. 2012. Correcting the imbalance of the World Heritage List: Did the UNESCO strategy work? International Organizations Studies 3(1):25-40.

Swiderska, K., Y. Song, J. Li, R. Pant, S. Vedavathy, D. Mutta, P. Munyi, H. Herrera, and A. Argumedo. 2009. Protecting community rights over traditional knowledge: implications of customary laws and practices. Key findings and recommendations 2005-2009. International Institute for Environment and Development, London, UK.

Titchen, S. M. 1996. On the construction of ‘outstanding universal value’: some comments on the implementation of the 1972 UNESCO World Heritage Convention. Conservation and Management of Archaeological Sites 1(4):235-242. https://doi.org/10.1179/135050396793138971

Tjuatja, L., V. Chen, T. Wu, A. Talwalkwar, and G. Neubig. 2024. Do LLMs exhibit human-like response biases? A case study in survey design. Transactions of the Association for Computational Linguistics 12:1011-1026. https://doi.org/10.1162/tacl_a_00685

Törnberg, P. 2024. Large language models outperform expert coders and supervised classifiers at annotating political social media messages. Social Science Computer Review. https://doi.org/10.1177/08944393241286471

UNESCO. 2011. Future of the World Heritage Convention. 18th session of the General Assembly of States Parties (7-9 November 2011). WHC-11/18.GA/11. UNESCO World Heritage Centre. Paris, France.

UNESCO. 2023. Operational Guidelines for the Implementation of the World Heritage Convention. WHC.23/01. UNESCO World Heritage Centre. Paris, France.