Audiovisual preservation in the television domain constitutes an increasing challenge in the digital society. The massive production of content, combined with technological obsolescence and storage limitations, demands increasingly rigorous selection criteria. Within this context, it becomes necessary to rethink traditional archival methodologies and to explore new approaches to archival appraisal (Taurino & Aitaki, 2025b). Preservation prioritisation can be more clearly informed by archival appraisal criteria and future-oriented archival judgment.

Television archives accumulate materials that cannot be preserved in their entirety (Gimenez Rayo, 2007; Taurino & Aitaki, 2025b). Resource management requires prioritising which documents may become relevant for future collective memory. This decision is not neutral: It involves cultural, historical, and social value judgments that remain conditioned by the subjectivities of the present. Determining which audiovisual pieces may acquire historical significance thus becomes an act of prospective appraisal with profound implications (Gimenez Rayo, 2007; Taurino & Aitaki, 2025b).

Archival science has developed solid criteria for archival appraisal, and these have long combined retrospective and prospective judgments, even if the anticipatory dimension has not always been made explicit in operational terms. Documents are preserved or discarded based on their immediate utility or their proven significance over time. However, in a scenario of audiovisual overproduction, it is insufficient to limit decision-making to the recognition of already consolidated values. Analytical tools are needed that help archivists examine possible future significance now of creation or management without reducing appraisal to a purely technical exercise (Schellnack-Kelly & Modiba, 2024).

For the purposes of this article, acquisition, appraisal, selection, and preservation prioritisation are treated as related but distinct archival functions. The proposal does not reduce appraisal to acquisition, nor does it claim that technical preservation constraints determine archival value. Rather, it examines how appraisal-informed preservation prioritisation may help archival professionals decide which television materials require earlier, deeper, or more resource-intensive preservation actions under conditions of audiovisual overproduction and technological vulnerability.

AI offers a broad framework of possibilities for such an approach. Machine learning systems can detect patterns in large datasets, identify cultural trends, and recognise contextual elements that humans can hardly process in a comprehensive way (Wu et al., 2023). Applying these capacities to audiovisual preservation opens the possibility of designing decision-support systems capable of identifying signals of potential archival significance before such significance becomes historically consolidated (Colavizza et al., 2022). Yet these capacities should be understood as analytical support for archival appraisal rather than as autonomous mechanisms for assigning value.

In the television field, this opportunity becomes even more evident. Television, as a mass medium, reflects both historical events and the cultural and social transformations of each era. Documents that appear trivial today may later become essential testimonies of collective experience. The challenge lies in discerning, in the present, which of them may prove significant in future interpretive contexts. Incorporating AI into this process does not eliminate the human dimension of decision-making, but it can provide analytical support that complements archival expertise (Jaillant & Rees, 2023; Taurino, 2025a).

This article proposes a theoretical framework for AI-assisted predictive appraisal and preservation prioritisation. The concept is defined as the application of AI models to the anticipatory appraisal of televised audiovisual documents. The proposal builds upon the convergence of heritage preservation studies, machine learning, and computational cultural analysis. By articulating these perspectives, it presents a conceptual model that combines traditional archival criteria with emerging technological tools.

The article also addresses the epistemological and ethical implications of supporting value-related archival judgments prospectively. The use of algorithms to inform decisions about preservation priorities introduces dilemmas regarding neutrality, bias, and transparency. Reflecting on these dimensions is essential to avoid the risk that AI reproduces cultural inequities or obscures minority narratives.

The originality of this proposal lies not in claiming that archives have never anticipated possible future values, but in articulating a more explicit conceptual framework for how AI may support future-oriented appraisal and preservation prioritisation in television audiovisual archives. This approach does not seek to replace human judgment, but rather to enrich it with new analytical methods. In this sense, the article seeks to contribute to archival discussions on appraisal, accountability, and human oversight under conditions of archival abundance.

The main objective of this work is to propose a theoretical framework for AI-assisted predictive appraisal and preservation prioritisation that explores how AI can be used to support future-oriented assessment of the potential relevance of televised audiovisual documents. Rather than testing a formal hypothesis, the article is guided by the problem of how archival, contextual, and technical-risk considerations may be articulated in a future-oriented appraisal framework for television documents. In this sense, the study seeks to clarify the conceptual conditions under which such a framework might be discussed and later examined in applied settings.

The objects of study are television documents understood as records of high informational, cultural, and social density. These materials not only fulfil functions of communication and entertainment but also serve as historical testimonies of a time and a society. Television has documented conflicts, political transformations, cultural expressions, and technological changes, generating a heterogeneous heritage whose preservation is essential. The difficulty lies in identifying, within this massive stream of production, which documents may acquire archival or historical significance in future interpretive contexts, without yet having the perspective of time.

The specific objectives of this study are oriented toward three dimensions. First, to analyse the theoretical foundations of archival appraisal in audiovisual archives, with an emphasis on television. Second, to explore the potential of AI as a decision-support tool for predictive appraisal. Third, to propose a conceptual model of AI-assisted predictive appraisal and preservation prioritisation that integrates these elements into an innovative archival framework. Achieving these objectives will enable progress toward a conceptualisation of archival appraisal that does not depend exclusively on retrospective criteria.

From a methodological perspective, the study adopts a qualitative, theoretical, analytical, and exploratory approach based on a critical review of specialised literature. The review is conceived as a critical conceptual literature review rather than as a systematic review or an empirical validation design. The research draws on academic sources indexed in reference databases such as Web of Science and Scopus, along with recent contributions available through Google Scholar. Sources considered include works from audiovisual archiving, digital preservation studies, AI applied to media, and epistemological approaches to cultural memory. This strategy allows the construction of an interdisciplinary framework that articulates archival tradition with emerging technological innovations.

The bibliographic search was conducted using terms in both Spanish and English, given the international nature of the field. Among the descriptors employed were: “inteligencia artificial” AND “archivos audiovisuales” AND “preservación”; “television archives” AND “AI-based prediction” OR “documentary appraisal”; “digital preservation” AND “cultural heritage” AND “machine learning”; as well as combinations integrating “archival valuation,” “predictive models,” or “prospective curation.” The aim of this strategy was to gather both technical and humanistic perspectives that enrich the analysis from different academic traditions.

Selection of the corpus was guided by thematic relevance, conceptual usefulness, disciplinary representativeness, and the need to connect archival appraisal literature with television archives, AI applications, and debates on bias, plurality, and human oversight. The review was not intended to be exhaustive in a systematic sense; rather, it prioritised representative contributions in four intersecting areas: archival appraisal theory, television audiovisual archives, AI applications in archival and heritage contexts, and debates on accountability, bias, and human oversight.

The analysis of the literature goes beyond descriptive synthesis, functioning instead as an exercise in problematisation. Points of convergence and divergence between traditional archival science and AI approaches are examined in order to delineate a novel conceptual space. From this review, hypothetical scenarios of application are outlined, in which AI systems identify patterns in television content, context, and production metadata. These scenarios serve as analytical exercises to visualise how AI-assisted predictive appraisal could operate in practice and what ethical and epistemological risks might emerge from its implementation.

The methodology also emphasises that this study does not include empirical fieldwork or technical algorithm development but rather focuses on the theoretical construction of a conceptual model. The prospective character of the approach requires a speculative analysis grounded in existing literature rather than immediate empirical evidence. Nonetheless, the study acknowledges the necessity for future applications of AI-assisted predictive appraisal to integrate human validation. The archivist’s professional judgment remains irreplaceable for ensuring cultural relevance, interpretative plurality, and the mitigation of algorithmic biases.

The adopted methodology seeks to generate a robust framework that combines critical literature review, conceptual analysis, and hypothetical application scenarios. The dialogue between disciplines such as archival science, data science, and cultural studies is indispensable for assessing the potential and the limits of AI-assisted predictive appraisal. In this way, the methodological proposal not only delineates an emerging field but also provides tools to think critically about how and why AI may reconfigure the criteria for the preservation of televised audiovisual heritage in the future.

Archival appraisal has become a central concern of modern archival theory, particularly in twentieth-century debates on selection, evidential and informational value, documentation strategy, and macroappraisal. Earlier custodial approaches, exemplified by Jenkinson (1922), foregrounded authenticity, custody, administrative responsibility, and the evidential integrity of records rather than the archivist’s active selection on behalf of future researchers. Later contributions by Schellenberg (1956), Samuels (1986), Booms et al. (1987), and Cook (2005, 2011) more directly address appraisal as a situated judgment about value, evidence, representativeness, and future use.

Classical and contemporary literature on appraisal has emphasised that archival systems cannot preserve all records without some form of selection, although the theoretical basis for that selection differs across archival traditions. In Schellenbergian and post-Schellenbergian approaches, overabundance of information, combined with limitations of space, resources, and technology, necessitates appraisal frameworks that ensure the sustainability of archival systems (Cook, 2005; Schellenberg, 1956). More recent debates on documentation strategy and macroappraisal reinforce the idea that selection is inseparable from representational choices, social context, and the construction of future memory (Cook, 2005, 2011; Samuels, 1986).

This perspective is especially relevant here because it shows that future-oriented judgment is not foreign to modern appraisal theory, while also requiring caution against any objective or deterministic conception of value. Booms et al. (1987) are particularly useful in this respect because they connect appraisal to socially situated judgments about documentary heritage and warn against treating value as universal, ahistorical, or fixed. What contemporary AI brings is not the invention of anticipation, but the possibility of making some inferential processes more explicit, scalable, documented, and reviewable in environments of archival abundance.

This line of argument also connects the proposal with documentation strategy and macroappraisal. Like those approaches, AI-assisted predictive appraisal assumes that selection is socially situated and that representational gaps must be addressed explicitly rather than left to accumulative logics. At the same time, the framework does not replace provenance with thematic pertinence; instead, it seeks to articulate provenance-based archival control with contextual and thematic indicators that may help archivists examine future-oriented appraisal judgments under conditions of archival abundance.

Television introduces unique specificities into this debate. Audiovisual production for television is characterised by its massive scale, its diversity of genres, and its strong dependence on technology. Television archives face the challenge of selecting, from an immense flow of materials, which should be preserved in the long-term (Kramp, 2019). News broadcasts, entertainment programs, debates, documentaries, and live events coexist in the same environment, creating a complexity greater than that of many other archival repositories.

Studies on television archives have also highlighted the impact of technological obsolescence (Taurino & Aitaki, 2025b). Magnetic and analogue formats employed for decades present risks of physical deterioration and inaccessibility due to the lack of playback devices. Digitisation has extended the life of these documents but has exponentially multiplied the volume of materials to be managed (Nachreiner, 2009). The paradox is clear: The easier it becomes to produce and store, the more difficult it becomes to select and preserve.

In this context, archival appraisal and preservation prioritisation in television archives must combine archival, cultural, and political criteria with technical risk, urgency, and feasibility factors. Decisions about which programs to preserve are not determined solely by their aesthetic quality or informational relevance but also by factors such as the representativeness of social voices, their capacity to testify to historical events, their potential to become cultural memory, and their vulnerability to loss or inaccessibility (Stockinger, 2025). Selection is never neutral and, as several authors have underlined, always reflects a situated perspective on what should be remembered. This is also where justice, plurality, and minority representation become archival concerns rather than merely external ethical add-ons (Caswell et al., 2017; Cline, 2021; Harris, 2002; Punzalan & Caswell, 2016; Schwartz & Cook, 2002;).

Digital preservation has attempted to address these challenges through strategies of migration, normalisation, and metadata management. The emphasis has been placed on ensuring long-term accessibility and technical stability, especially in environments marked by format obsolescence and infrastructure dependency. These approaches are indispensable, but they do not by themselves resolve the question of archival value. In other words, technical constraints may shape preservation urgency and feasibility without determining value in the archival sense.

Some international projects have experimented with automated prioritisation systems. These initiatives generally rely on algorithms that detect duplication, define levels of technical quality, or highlight the rarity of certain records. While useful, such tools do not address the cultural and social dimensions of archival relevance. Audiovisual preservation is not exhausted by maintaining technical copies but also involves recognising which documents embody narratives, memories, and sensibilities that may later be interpreted as significant for future generations (Pendergrass et al., 2019).

The emergence of AI is reconfiguring this debate by introducing capacities for large-scale analysis and automated learning. Machine learning models can process audiovisual corpora, identify thematic recurrences, recognise entities, and detect contextual or formal patterns that might be relevant to archival decision-making. Yet these capacities should be understood as analytical supports for appraisal, not as autonomous mechanisms for assigning value. Their usefulness depends on how they are embedded within archival criteria, human interpretation, and accountable workflows.

Applications of AI in audiovisual archives have advanced along several fronts. Developments include automatic cataloguing of images and sounds, facial recognition, and text analysis in subtitles and scripts. Experiments have also been conducted with semantic classification systems that facilitate indexing and retrieval. These experiences demonstrate that AI can optimise technical processes in archival management, although they have rarely been framed from a predictive perspective (Wu et al., 2023).

An emerging area is computational cultural analysis, which explores how algorithms can detect narrative and aesthetic trends within large audiovisual corpora. This approach has been employed in social media studies and digital media analysis but has not yet been systematically applied to television archives. The potential extrapolation of its methodologies suggests fertile ground for predictive preservation, as it enables the identification of cultural significance patterns that could become markers of cultural heritage significance in the future (Roy et al., 2024).

Critical literature on AI nonetheless warns of epistemological and ethical risks. Dependence on algorithms trained on large-scale datasets may reproduce cultural and social biases (Chen, 2025). The invisibilisation of minorities, overrepresentation of dominant discourses, and normalisation of hegemonic narratives are well-documented risks. Applied to audiovisual preservation, such biases could consolidate a partialised memory that reinforces cultural inequalities (Breemen & Breemen, 2025).

Another relevant debate concerns algorithmic transparency and interpretability. Deep learning systems often function as black boxes, with decisions that are neither easily comprehensible nor auditable (Doshi-Velez & Kim, 2021). This stands in tension with archival principles, which demand traceability, accountability, and clear justification of appraisal decisions (Franks, 2023). For predictive preservation to be viable, it must incorporate mechanisms that allow professionals to understand and supervise algorithmic recommendations.

Ethical studies on AI applied to heritage also emphasise the importance of human validation. No automated system can replace the culturally situated perspective of the archivist. The risk of full delegation to algorithms is that preservation decisions may devolve into purely technical operations, devoid of attention to social and historical contexts. For this reason, predictive preservation models should be conceived as analytical supports, not substitutes for professional appraisal. Recent work on InterPARES Trust AI, Records in Contexts, and paradata-based accountability further underlines the importance of documenting both the records and the processual conditions under which AI-assisted inferences are produced (Cameron et al., 2023; Feliciati, 2023; Feliciati & Duranti, 2025; International Council on Archives, 2023).

Within archival science, concepts such as predictive appraisal or digital curation have been proposed to address the need to incorporate future-oriented criteria into archival management. However, these notions have not yet been systematically articulated in relation to AI. The present article therefore adopts predictive appraisal as its primary archival frame and treats predictive preservation as a preservation-oriented formulation within that wider discussion. The idea of predictive archival science’ has only begun to appear tentatively in some studies, but it lacks robust conceptual development and specific applications to the television domain.

At this juncture, a preservation-oriented reformulation of predictive appraisal becomes pertinent. The term seeks to synthesise the convergence of the archival tradition of appraisal, the specific challenges of audiovisual television preservation, and the potential of AI. By articulating these three axes, the proposal offers a conceptual framework capable of addressing the gaps identified in the literature.

The state of the art thus reveals a set of unresolved tensions. Television archival practice requires more refined selection and prioritisation criteria than those currently in place. Digital preservation has solved technical problems but not epistemological ones. AI provides powerful tools, yet it introduces risks of bias and opacity. What is lacking is a model that integrates these dimensions into a coherent theory of predictive appraisal, oriented toward preserving materials that may acquire significance in future interpretive contexts.

Consequently, the literature review highlights both the richness and the limitations of current approaches. Solid traditions of archival appraisal exist, but they have not always been translated into explicit frameworks for AI-assisted future-oriented appraisal in television archives. Digital preservation techniques have been developed, but with limited attention to cultural anticipation. AI has been applied to archives, but without a systematic focus on potential future relevance. The resulting line of inquiry is therefore not a break with archival tradition, but a conceptual attempt to connect appraisal theory, audiovisual preservation, and computational analysis under conditions of human oversight and archival accountability.

The analysis undertaken makes it possible to outline a set of conceptual and technical strategies for discussing AI-assisted predictive appraisal and preservation prioritisation in television audiovisual archives. Rather than presenting empirically demonstrated results, this section organises an initial conceptual proposal for how such an approach might be structured in practice. The proposal is structured around three fundamental axes, content, context, and metadata, which, when integrated, constitute an analytical framework for examining possible trajectories of future historical relevance. The proposed appraisal dimensions and their relationship to AI-assisted signal, evidentiary limitations, and human validation are summarised in table 1.

From the perspective of archival appraisal theory, the proposal is closer to documentation strategy and macroappraisal than to a merely thematic reordering of collections: It retains provenance as an archival principle while using contextual and thematic indicators only as aids to appraisal and preservation prioritisation.

The proposal for AI-assisted predictive appraisal and preservation prioritisation in television audiovisual archives should be understood as a conceptual extension of archival appraisal rather than as a radical break with it. While modern appraisal theory has included forms of prospective reasoning about future use and significance, the incorporation of algorithms capable of assisting anticipatory judgment introduces new analytical possibilities together with new epistemological and ethical risks. Examining which content may acquire significance in future contexts can optimise resources and support the conservation of high-potential materials, but it may also constrain the interpretative breadth inherent to collective memory by privileging patterns detectable by AI over emergent phenomena or unexpected cultural ruptures.

From a theoretical perspective, the notion of archival value is neither universal nor fixed. Audiovisual documents acquire meaning within specific social, historical, and cultural contexts, and their appraisal depends on the interaction between the agents who produce, consume, and preserve them. A predictive model based solely on algorithmic analysis could inadvertently fix criteria that favour certain narratives, reproducing historical or cultural biases present in the training data. This underscores the necessity for AI-assisted predictive appraisal to integrate human supervision mechanisms that ensure interpretative plurality and prevent systematic exclusions.

In this regard, human judgment remains indispensable. AI can assist in identifying patterns relevant to appraisal, but it cannot replace critical reflection on the pertinence, ethics, or context of documents. The complementarity between technology and human judgment thus becomes a methodological principle: Automation enhances analytical capacity, while the archivist retains authority over the final definition of relevance and interpretation of results.

Another central point of discussion is the tension between anticipation and openness to the unexpected. Predictive preservation cannot, and should not, eliminate the possibility of conserving documents that may not seem relevant at present but could acquire significance in future contexts due to social, cultural, or political changes. Algorithmic planning must coexist with inclusive and flexible preservation strategies that respect the uncertainty inherent in cultural evolution.

Furthermore, cultural and community dimensions are relevant without displacing professional responsibility. Possible future archival values cannot be defined solely from a technological or institutional perspective; in certain cases, especially those involving underrepresented communities or culturally sensitive records, it may be necessary to incorporate forms of consultation or plural validation. Final responsibility, however, remains with archival institutions and professionals, who must document the criteria, evidence, and decisions applied.

Finally, this analysis highlights clear research horizons: Exploring the integration of predictive archival practices with digital humanities, evaluating real-world application scenarios, and designing mechanisms for the periodic review of automated decisions. Open questions arise regarding the fate of discarded documents and how to ensure that future preservation reflects a plurality of memories, avoiding algorithmic homogenisation of audiovisual history. Future work should also address paradata, auditability, documentation of model versions, decision logs, human overrides, and hybrid workflows in specific archival settings.

The article suggests that AI-assisted predictive appraisal and preservation prioritisation in television audiovisual archives constitute a promising conceptual approach for thinking about complex collections characterised by high informational, narrative, cultural, and technical density. The study argues that audiovisual documents do not merely transmit objective information; they also construct social meanings and may acquire archival significance across changing historical contexts. In this sense, AI can support decision-making regarding conservation, access, and curation, if it remains subordinated to professional judgment and documented human oversight.

The theoretical analysis situates predictive preservation within an interdisciplinary framework where audiovisual archiving, archival appraisal, digital preservation, and computational analysis converge. This integration allows AI to be understood not as an autonomous evaluator of archival value, but as a decision-support instrument capable of identifying patterns, absences, correlations, and priority signals that require professional interpretation.

From a methodological perspective, the critical review of the specialised literature indicates that, although advanced technical developments exist in audiovisual content analysis and automatic tagging, there is still a scarcity of studies that explicitly integrate the ethical, cultural, and epistemological dimensions of future-oriented appraisal and preservation prioritisation. This gap highlights the need to propose hybrid models in which automation and human supervision coexist, ensuring inclusive and pluralistic criteria of archival value. The consideration of hypothetical scenarios and reflection on the risks of algorithmic bias provide a conceptual framework to anticipate future challenges and guide more responsible preservation practices.

The discussion of ethical and cultural implications emphasises the necessity of maintaining the possibility of preserving documents that may not currently be deemed relevant but could acquire significance due to social change, historical reinterpretation, or emerging research interests. The model therefore rejects any definitive or objective prediction of value and instead frames AI outputs as provisional, situated, reviewable, and accountable signals.

In practical terms, the study proposes that AI integration in television audiovisual archives should focus on enhancing accessibility, strategic curation, metadata enrichment, risk awareness, and retrieval of audiovisual records without compromising historical integrity or interpretative richness. The analysis points toward a more proactive archival model capable of supporting preservation planning and identifying materials that deserve professional review, without delegating authority over collective memory to technology.

AI-assisted predictive appraisal and preservation prioritisation emerge as promising tools for audiovisual archive management, provided they are implemented under ethical, critical, and methodologically sound criteria. The combination of algorithmic analysis, archival criteria, processual documentation, and human judgment offers a pathway to enrich description, access, and curation of television documents, contributing to a more inclusive and plural representation of audiovisual memory. Future research should test the proposed framework in controlled archival settings, develop auditable workflows, and examine how predictive models can be integrated with professional standards, community-sensitive appraisal, and long-term preservation strategies.

This research was funded by the Junta de Extremadura and the ERDF Funds “A way of making Europe” [Research Group QUINARI SEJ013 GR18130].