WSIS Action Line C3 Access to information and knowledge

The Moderator outlines that access to information is a multifaceted issue encompassing legal frameworks, digital infrastructure, connectivity, and normative instruments. Monitoring how member states are enabling or creating instruments for access to information requires attention to many interconnected dimensions. These include linguistic diversity, data governance, privacy, transparency, and participation.

The Moderator referenced UNESCO's custodianship of SDG indicator 16.10 and research on world trends in freedom of expression and access to information laws . The Moderator also highlighted the interconnection between access to information and dimensions such as linguistic diversity, data governance, and privacy , as well as the role of digital infrastructure and connectivity gaps .

The Moderator observes that discussions about access to information tend to focus on digital connectivity and overlook the physical dimension of access, such as the built-environment barriers that prevent people with disabilities from entering libraries or other information spaces. This is a significant gap in how access to information is conceptualised and discussed in forums such as WSIS. The Moderator acknowledges that connectivity is also physical but notes that built-environment accessibility is rarely addressed.

The Moderator noted that access to information is often thought of as a non-physical subject, and that while connectivity is discussed as a physical issue, physical accessibility in the built-environment sense is rarely addressed in these discussions .

Dr von Franqué argues that simply counting access to information laws is too narrow a measurement approach. She proposes including literacy indices, copyright laws, and freedom of panorama restrictions as additional indicators, since these also affect people's practical ability to access and share information. Barriers such as restrictions on photographing public statues illustrate how copyright regimes can limit access to knowledge.

She gave the example of freedom of panorama, noting that in some places it is not permitted to photograph a public statue, which limits access to visual information and knowledge . She also suggested that Wikimedia Commons, which allows uploading of knowledge reports and other materials, faces barriers that could serve as useful indices for access to knowledge .

Dr von Franqué warns that AI bots are aggressively scraping Wikimedia's infrastructure in ways that resemble DDoS attacks, hiding their identity and causing significant resource costs. This behaviour is harming the digital commons in a substantial way, consuming around 64% of Wikimedia's infrastructure resources. The scale of the problem is illustrated by the figure of 1.3 billion bot attacks per minute.

She stated that AI bots are scraping Wikimedia infrastructure and behaving like DDoS attacks, costing approximately 64% of infrastructure resources . She noted that Wikimedia fights 1.3 billion bot attacks every minute, and that the situation was only manageable because a small Canadian company provided open-source bot protection at a low cost .

Dr von Franqué draws a distinction between large language models (LLMs), which are based on statistical outputs rather than factual accuracy, and symbolic AI such as Wikidata, which is precise and scientifically verifiable. She argues that LLMs are therefore unreliable for knowledge dissemination, particularly across languages. Symbolic AI, by contrast, offers a more trustworthy foundation for structured knowledge.

She noted that Wikidata can be described as a symbolic AI that is feasible by scientific standards, whereas large language models are based on statistics and do not prioritise correct outcomes . She also referenced a session where automatic translation by an LLM was very poor even for French, and that Russian was equally basic, illustrating language reliability issues .

Dr von Franqué acknowledges that AI, including large language models, can be useful for data and information collection and is improving in translation. However, she emphasises that the quality of any AI system depends fundamentally on the quality and structure of its underlying data. Libraries, as the oldest knowledge-structuring institutions, are therefore a critical foundation for effective AI.

She stated that the basis of every good AI, including large language models, is good structured data and good structured knowledge, and that libraries are the oldest knowledge-structuring institutions . She suggested starting with basic library metrics such as number of libraries and distance to the nearest library as a foundation for layering more complex access indicators .

Dr von Franqué proposes that libraries offer a practical and measurable baseline for assessing access to knowledge. Simple metrics such as the number of libraries, the distance a person must travel to reach one, and the resources available within them can be established first. These can then be layered with more complex indicators such as literacy indices and educational access measures.

She suggested counting libraries, measuring the distance to the nearest library, and assessing the number of books available as basic starting points for access to knowledge measurement . She also referenced the use of literacy and media literacy indices to assess the impact of libraries on education, noting that a library that is closed all year will have no educational impact .

Dr von Franqué proposes that Wikidata can serve as a practical tool for systematically recording and querying accessibility features of libraries and other institutions. By entering data on whether a library has wheelchair access or hearing-impaired support, it becomes possible to generate indices showing what percentage of libraries are accessible. This turns Wikidata into a useful measurement instrument for accessibility.

She explained that libraries could be entered into Wikidata with fields indicating whether they are accessible for wheelchair users or have hearing-impaired support, and that this data could then be queried to produce accessibility indices . She also noted that librarians themselves could contribute by adding accessibility information to Wikipedia articles, which Wikidata could then draw upon .

Dr von Franqué argues that existing digital public goods, particularly Wikimedia projects such as Wikipedia and Wikidata, are underutilised within the WSIS framework. These tools are already recognised by the United Nations as digital public goods and offer significant infrastructure for open knowledge. She suggests they should be more deliberately integrated into WSIS action lines rather than overlooked.

She noted that Wikipedia and Wikidata are both recognised as digital public goods by the United Nations . She expressed that it feels awkward to highlight this as though she were a salesperson, but emphasised that these projects already exist in 350 languages and represent global assets that are being left out of the picture in WSIS discussions .

Dr von Franqué raises concern that governments worldwide are channelling financial resources into building data centres and investing in AI, rather than supporting existing open knowledge infrastructure such as Wikimedia projects. This shift in funding priorities poses a financial threat to the digital commons. The implication is that public investment is flowing towards proprietary or state-controlled AI rather than shared, open resources.

She stated that governments all over the world are collecting money to build data centres and invest in AI, identifying this as a financial challenge for digital commons projects .

Dr von Franqué highlights that Wikimedia projects, including Wikipedia, already operate in 350 languages, making them a significant multilingual infrastructure for global knowledge access. This positions them as a practical tool for bridging language-based information gaps. She also notes that local communities can contribute their own knowledge to these platforms, including in their own languages.

She stated that Wikimedia projects already exist in 350 different languages . She also noted that Wikipedia allows users to add local knowledge, such as information about local snake species or library accessibility, and that Wikidata can store structured information from libraries across countries .

Ms Dixon argues that the drive to standardise measurement of access to information and education may actually obscure the most effective interventions, which are those tailored to specific local contexts. The most meaningful difference in her programme's work comes from unstandardised, contextualised approaches. This creates a tension between the desire for comparable metrics and the reality of what produces impact.

She gave the example of community health worker training, noting that what differentiates a course in Jordan from one in Kakuma is that public health concerns and priorities differ significantly between settings . She stated directly that the part of the course that makes the most difference is the tailored part, and that it is in unstandardising that they make the difference .

Ms Dixon explains that the InZone programme at the University of Geneva provides higher education to refugees in camps and remote settings, and that the most effective element of this provision is the tailoring of content to local needs. Different refugee settings have different public health priorities and concerns, meaning that a one-size-fits-all curriculum is insufficient. Contextualisation is therefore essential to meaningful access to education.

She described how community health worker training in Jordan differs from that in Kakuma because public health concerns and prevention priorities are not the same . She noted that the programme uses hybrid learning methods including Moodle, but that the in-person workshops with local actors implementing contextually relevant knowledge make the most difference .

Ms Dixon argues that communities in refugee and displaced settings should be participants in producing educational content, not merely recipients of externally designed curricula. Communities have specific knowledge of their own needs and priorities that outside institutions cannot replicate. This requires building a notion of useful knowledge that is defined in part by the communities themselves.

She gave the example of community members identifying the need to understand local snake species and how to prevent children from being bitten, as a priority for health risk prevention training . She noted that courses accredited by the University of Geneva still incorporate content proposed by the communities themselves, as they know what they need to know .

Ms Dixon highlights that refugees often have highly disrupted educational histories, having moved across different educational systems, languages, and regions, and sometimes losing documentation of their qualifications. Recognising these diverse and non-linear pathways is essential for enabling meaningful access to higher education and knowledge. She references ongoing discussions with UNESCO about a qualification passport as one potential mechanism.

She referenced discussions with UNESCO about a qualification passport as a tool for recognising prior learning and disrupted educational pathways . She noted that refugees may have moved across multiple systems and languages, lost documentation, or been unable to sit exams despite completing training, all of which complicate recognition of their prior learning .

Ms Dixon argues that education in critical thinking and information literacy is one of the most important but underemphasised tools for combating misinformation. In refugee settings, where political sensitivities are heightened and communities may be exposed to conflict-related disinformation, the ability to critically evaluate information is particularly vital. She notes that preparation programmes should develop this capacity alongside technical knowledge.

She noted that refugee settings can be politically sensitive, with conflicts sometimes crossing borders into camps, making information integrity a major issue . She described preparation programmes aimed at developing critical thinking and the capacity to compare and evaluate information, noting that classical secondary education often teaches by rote without developing these methodological skills .

Ms Dixon argues that automatic translation is itself a significant source of unintentional misinformation, because the quality of machine translation is often so poor that it distorts the meaning of official communications. She illustrates this with a concrete example from a WSIS session where a French speaker's words were rendered nonsensically in English. This is not merely a quality problem but an ethical one, as it means the actual content of official statements is lost.

She described an incident at WSIS where a country representative speaking French had their words automatically translated into English in a way that was nonsensical; the phrase 'statistique des abonnés' was translated as 'Japanese statistics' . She noted that at the WSIS session, automatic translation was only available for English, effectively disregarding the right of French-speaking country representatives to be understood .

Ms Dixon argues that the failure to provide quality translation at official international forums such as WSIS disregards the rights of non-English speakers to participate meaningfully. When automatic translation is only available for English, speakers of other official languages such as French are effectively marginalised. This is both a quality problem and an ethical one, as it undermines equal participation in global policy discussions.

She noted that at WSIS, country representatives are allowed to speak in French but automatic translation is only provided for English, which she described as disregarding their right to speak in French . She gave the concrete example of a French speaker's words being mistranslated as 'Japanese statistics', illustrating how poor translation quality renders official communications meaningless .

Speaker 1 warns that AI-generated content is polluting the information environment at a massive scale, citing data suggesting that 65% of social media conversations are produced by bots. This creates a situation where people believe they are interacting with humans when they are not, fundamentally distorting public discourse. The speaker frames this as a very serious problem that needs to be addressed in governance frameworks.

Speaker 1 cited Google data indicating that 65% of conversations on social media are produced by bots, describing this as very scary because it means people are talking to entities they believe are human but are not . The speaker referenced the European Union's democracy shield initiative, which aims to create trusted European platforms where identification of persons is mandatory through the EU ID wallet, ensuring only humans act in that space .

Speaker 1 argues that AI interfaces are increasingly replacing human customer service agents but are largely ineffective, failing to understand users' needs and trapping them in unproductive loops. Companies that replace human workers with inadequate AI to save money should face sanctions. There should be a mandatory option to escalate to a human when AI interfaces fail.

Speaker 1 described personal experience of trying to change an address in a database and being unable to do so through an AI interface, losing significant time in the process . The speaker noted that 90% of the time AI interfaces do not understand what users want, and that in 90% of cases there is no option to escalate to a human, leaving users trapped in loops .

Speaker 1 raises the alarm that AI now makes it technically possible to alter digital archives, creating a serious risk to the integrity of historical memory. This could enable politically motivated actors to falsify historical records in ways that are indistinguishable from authentic documents. The speaker notes this is already reportedly occurring, with Ukraine cited as a country facing this problem.

Speaker 1 referenced a seminar with WIPO on archives, where the problem of AI-enabled alteration of archives was discussed . The speaker gave the example that it is now possible to make it appear that Stalin was a nice person, with no way to distinguish authentic from falsified records . Ukraine was cited as a country reporting this as a significant problem .

Speaker 1 argues that the General Assembly's decision to grant a permanent mandate to the IGF and extend the WSIS mandate creates a timely opportunity to align these processes with other UN frameworks, particularly the Global Digital Compact. Operating in isolation is no longer viable; the various UN processes must interact and align their deadlines and objectives. This alignment is necessary for WSIS to contribute meaningfully to broader UN goals.

Speaker 1 noted that the General Assembly gave a permanent mandate to the IGF and prolonged the WSIS mandate, describing this as an opportunity to align the process with other UN reflections . The speaker stated that the GDC implementation concerns all of the UN family and that WSIS must contribute to it .

Speaker 1 argues that WSIS must actively build bridges with other governance fora rather than operating in isolation, in order to ensure that its discussions on access to information have broader reach and impact. The IGF Policy Network for Digital Inclusion, which evolved from the Policy Network for Meaningful Access, is cited as an example of a forum with which WSIS should engage. Parallel actions and shared points of attention are needed across these bodies.

Speaker 1 noted that they co-chaired the IGF Policy Network for Meaningful Access, which has now become the Policy Network for Digital Inclusion with an enlarged scope, and argued that this broader perspective is important . The speaker called for creating bridges and parallel actions to bring together the same points of attention across different fora .

Speaker 2, a librarian from Ghana, highlights that physical barriers such as the absence of ramps and lifts prevent people with disabilities from entering library buildings in the first place. This is a fundamental but often overlooked dimension of access to information. Even where digital resources exist within libraries, they are inaccessible if people cannot physically enter the building.

Speaker 2 described libraries that have digital and library resources but no ramps or lifts for people with disabilities to access the building . The speaker framed this as a serious barrier to information access that needs to be addressed .

Speaker 2 points out that even when people with disabilities can physically access a library, the absence of assistive devices means they cannot meaningfully access the information within. Visually impaired users lack devices to read materials, and hearing-impaired users lack audio-visual aids. This represents a second layer of exclusion beyond physical access.

Speaker 2 noted that for visually impaired users, there are no assistive devices to help them read and access information, and for hearing-impaired users, there are no audio-visual aids to assist them . The speaker described this as a challenge particularly affecting visually and hearing-impaired people .

Speaker 2 raises the urgent need to capture and preserve indigenous knowledge, particularly in Africa, where vast amounts of knowledge about medicinal plants and other traditional practices remain undocumented. This knowledge is at risk of being lost if it is not systematically recorded and preserved. Ensuring its preservation is essential for future generations.

Speaker 2 stated that there is a huge amount of data on indigenous knowledge, particularly regarding medicinal plants in Africa, that has not been captured, and argued that there is a need to capture and preserve it for future generations .

Mr Engelson argues that the true gatekeepers of access to information in the digital age are search engines and AI platforms such as Google and ChatGPT, not the traditional internet governance bodies such as ISOC or ICANN. Despite this, these platforms are rarely discussed in measurement frameworks or governance debates. This represents a significant blind spot in how access to information is understood and governed.

He argued that the real governance of the internet is not ISOC or ICANN but Google and ChatGPT, and that this is rarely discussed . He used the metaphor of a large library where millions of books are available but inaccessible because users do not know which book to look for, contrasting this with small open-access libraries where browsing is possible, to illustrate the problem of navigating vast information .

Mr Engelson challenges the premise that access to information is still primarily a problem of scarcity, arguing that the real challenge today is navigating an overwhelming abundance of information. When a single search query returns 785 million results, the ability to find meaningful and relevant information is severely compromised. The nature of the access problem has fundamentally shifted from shortage to overload.

He noted that never in human history has there been so much access to information, and that even refugees have access to billions of documents through mobile phones . He gave the example of a search query returning 785 million answers, arguing that the issue of access is now less about availability and more about how to find meaningful information .

Mr Engelson makes a provocative argument that the dominance of English in international forums such as WSIS allows participants to build high-level careers without genuinely understanding what is being discussed. This undermines the quality and integrity of global policy discussions. He connects this to a broader critique of the lack of critical thinking and meaningful content in UN-level discourse.

He stated that participants can make whole high-level careers without really understanding what they talk about, and that everyone speaks only English . He also referenced a journalist's observation that media journalists do not attend WSIS, and cited a former UNESCO WSIS chairman who expressed puzzlement at this absence .