Social Media and Language

Chapter 136 examines social media as a linguistic institution rather than merely a communications technology. Facebook, WhatsApp, YouTube, Instagram, short-video platforms, messaging groups and emerging AI-assisted interfaces now sit between speech, writing, performance, migration and public debate. They make some language forms easy to record and circulate, others easy to search, and still others difficult to type or discover. The result is not a simple shift from “offline” to “online” language. Digital media reorganise the relative visibility of scripts, spellings, registers, accents, songs, jokes and political vocabularies while connecting speakers who may live in different districts, states or countries. The regional frame remains essential. Maithili, Angika and Bajjika do not occupy identical institutional positions, and Mithila, Vajji and Anga do not form one current administrative unit. India’s Census 2011 C- 16 table makes Maithili directly visible as a named mother-tongue category, while the Bihar table does not give Angika and Bajjika the same headline statistical visibility. Nepal’s 2021 census, by contrast, separately records Maithili and Bajjika in Madhesh Province. Platform data introduce another asymmetry: commercial services may know enormous amounts about users while releasing little language-specific regional evidence. This chapter therefore combines official language statistics, telecommunications evidence, national internet- use research and qualitative scholarship without inventing platform-specific counts for the historical-cultural region. 136.1 Social media is now an infrastructure of everyday language change Language change once depended heavily on face-to-face interaction, schools, print, broadcasting, migration and institutions. Social media does not replace those mechanisms; it connects them at much higher speed. A phrase spoken in a village can be recorded on a phone, circulated through a family group, remixed into a short video, commented on by migrants in Delhi or the Gulf and returned to the village with a new spelling or comic association. The crucial historical change is the compression of distance and delay. Linguistic innovations can travel through weak ties as well as kinship networks, and archived or searchable posts allow forms to remain visible after the original conversation has ended. Social media therefore operates as infrastructure: it changes the pathways through which language moves. 136.2 Platforms do not simply carry pre-existing languages Every platform has affordances and constraints. Text boxes reward what can be typed; search engines reward what can be spelled predictably; voice notes reduce the need for literacy; video privileges performance, facial expression and sound; recommendation systems rank some material above other material; monetisation rewards attention rather than linguistic representativeness. These design features affect language without mechanically determining it. Users adapt platforms, invent abbreviations, switch scripts, attach subtitles, use hashtags and circulate screenshots. The correct unit of analysis is therefore the relationship between platform design and social practice, not a claim that technology “causes” a single linguistic outcome. 13991399 GAJENDRA THAKUR Figure 540 — Social media and language form a feedback system: platform affordances shape visible forms, audience response and algorithmic selection, which feed back into community norms and everyday usage. 136.3 The regional language field is multilingual before it becomes digital Digital multilingualism in Mithila, Vajji and Anga begins with an already multilingual social world. Maithili, Bajjika, Angika, Hindi, Urdu, Nepali, Bhojpuri, English and other languages overlap through locality, schooling, administration, marriage, work and migration. Census of India C-17 explicitly treats bilingualism and trilingualism as measurable features, while Nepal’s census documents a highly plural mother-tongue landscape. A social-media post that mixes Maithili and Hindi, or Bajjika and Nepali, should not automatically be read as digital corruption of a previously pure code. It may represent an established repertoire made newly visible in writing. The historian must distinguish long-standing multilingualism from changes introduced by digital media. 136.4 Maithili has strong demographic presence but unequal digital resources across domains Maithili benefits from constitutional recognition in India, a large speaker base on both sides of the border and substantial literary production. Yet recognition does not guarantee equal digital support in keyboards, speech recognition, spell-checking, moderation, search, subtitles or machine translation. A language can be demographically large and still function as a low-resource language in particular computational systems. Social media partly compensates because video and audio allow creators to publish without sophisticated language technology. At the same time, discoverability often depends on metadata typed in Devanagari, Roman script, Hindi or English. The digital strength of Maithili is therefore domain-specific rather than uniform. 136.5 Angika and Bajjika reveal how digital visibility can diverge from official classification Angika and Bajjika demonstrate why institutional statistics cannot be treated as complete maps of linguistic life. In the Bihar Census 2011 C-16 presentation, neither appears with the same directly retrievable headline status as Maithili, although both have strong regional identities and public cultural use. Nepal’s 2021 census separately records Bajjika and shows its major presence in Madhesh Province. Online, users can name a language in a channel title, hashtag or video description even when official tables classify it differently. HISTORY OF MITHILA, VAJJI & ANGA — VOLUME II Social media thus creates a form of self-indexing: communities can assert names and boundaries in public metadata. That visibility is culturally important, but it is not equivalent to a census count or legal recognition. Figure 541 — Madhesh Province mother tongues in Census 2021: Maithili 41.73%, Bhojpuri 18.81%, Bajjika 18.44%, Nepali 5.76%, Tharu 4.17%, Urdu 4.08%, Tamang 1.65%, with the remainder grouped here as other languages. 136.6 Indic-language internet growth changes the economics of regional- language audiences The IAMAI–Kantar Internet in India Report 2024 estimated 886 million active internet users nationally, with rural India accounting for 488 million. It also reported that 98 percent of internet users accessed content in Indic languages and that 57 percent of urban users preferred regional-language content. These are India-wide figures, not Bihar estimates, but they establish an important market context. Regional-language content is no longer a marginal add-on to an English-dominant internet. Advertising, video distribution, news, entertainment and commerce increasingly have incentives to address users in non-English languages. For Maithili, Angika and Bajjika, the opportunity is real, but audience economics remain shaped by the much larger markets for Hindi and other major languages. 136.7 Video lowers the script threshold and gives oral vernaculars new public reach Video is especially important for language varieties whose speakers may be more confident in speech than in standardised writing. A singer, comic performer, farmer, teacher or political speaker can address an audience in a local variety without choosing a formal orthography for every sentence. Subtitles and titles can then be added in Devanagari, Roman script, Hindi or English. IAMAI–Kantar’s India-wide evidence shows video as the most common activity undertaken in Indic languages, ahead of social networking and search. For regional language history, this means that oral performance has become a major digital publication format. The public record of language is no longer dominated by books, newspapers and institutional broadcasting. 14011401 GAJENDRA THAKUR Figure 542 — India-wide Indic-language internet activities in the IAMAI–Kantar Internet in India Report 2024: video and music dominate, while social networking, shopping and search also occur substantially in Indic languages; these are not Bihar-specific estimates. 136.8 Voice messages return spoken registers to the centre of digital communication Messaging applications changed digital writing by making audio ordinary. Voice notes allow elderly users, people with limited literacy, migrants under time pressure and speakers unsure of standard spelling to participate in digital networks through speech. They preserve intonation, hesitation, local vocabulary and pronunciation that text messages flatten. Yet voice is also less searchable, less easily quoted and harder to archive systematically. A language may therefore thrive in private audio circulation while remaining underrepresented in searchable public text. Historians of digital language need to distinguish communicative vitality from archival visibility. 136.9 Devanagari has become the practical common script for much regional- language social media For Maithili, Angika and Bajjika, Devanagari is widely usable across phones, operating systems and mainstream platforms and overlaps with Hindi literacy acquired through schooling. This lowers the technical cost of publishing regional-language text. The same convenience can blur linguistic boundaries because identical script does not imply identical language. A Maithili sentence in Devanagari may be mistaken by automated systems or casual readers for Hindi unless vocabulary, metadata or context mark the distinction. Script convergence therefore supports access while making language identification more dependent on linguistic and social cues. 136.10 Roman transliteration is a functional digital register, not automatic evidence of language loss Roman-script writing appears frequently in messaging, comments and short captions because keyboards are familiar, code-switching with English is easy and users can avoid uncertainty about Devanagari spelling. Such writing is often unstable: the same word may appear in several Roman spellings. That instability reduces searchability and can make archiving difficult, but it also records pronunciation and informal style. Treating Roman transliteration simply as linguistic decline misses its practical function. The more useful HISTORY OF MITHILA, VAJJI & ANGA — VOLUME II question is whether users can move between scripts and domains when required—for example, between a Roman-script chat, a Devanagari public post and a formal Maithili publication. 136.11 Tirhuta or Mithilakshar has a different digital role: heritage, expertise and identity Unicode encodes Tirhuta at U+11480–U+114DF and identifies it as the traditional writing system of Maithili. Digital encoding makes genuine text possible rather than image-only reproduction, yet encoded existence is not the same as effortless everyday use. Fonts, keyboards, rendering support, user knowledge and platform compatibility still matter. On social media, Tirhuta often functions as a marker of heritage, scholarship, calligraphy and regional identity rather than as the default script for rapid conversation. This specialised visibility can nevertheless be historically significant: a script that had become difficult to reproduce in ordinary print can circulate globally as searchable text when the technical chain works. 136.12 Unicode, fonts and keyboards are part of language infrastructure A user can only type what the input system makes practicable and only read what the device can render. Unicode standardisation solved the fundamental problem of assigning stable code points, but practical inclusion requires more: fonts, keyboard layouts, search indexing, line breaking, text-to-speech, optical recognition and developer support. Chapter 116 treated Unicode and Tirhuta as technological recognition; the present chapter follows the social consequence. When technical support is uneven, users route around it through screenshots, images, Roman transliteration or another script. These workarounds preserve communication but weaken machine readability and long-term digital preservation. 136.13 Code-switching is a resource for audience design A single post may combine a Maithili or Bajjika spoken base with Hindi for a wider regional audience and English for technical terms, hashtags or prestige. Such code-switching can index education, humour, urbanity, migration or group membership. Creators also adjust language by platform: a family message may use a local variety, a public caption Hindi, and a professional profile English. Networked multilingualism therefore involves strategic audience design. The same person can inhabit several linguistic publics without abandoning a mother tongue. Analysis should ask who is being addressed and what social work each language performs. 136.14 Orthographic variation becomes publicly visible and therefore contestable Print institutions historically reduced variation through editors, proofreaders and house styles. Social media exposes far more unedited writing, including local pronunciations, spellings, abbreviations and hybrid forms. Comment threads can then become sites of correction: users debate whether a word is “proper” Maithili, whether a form is Hindi-influenced, or whether a Bajjika or Angika expression belongs to another label. This can strengthen metalinguistic awareness, but it can also convert normal variation into status conflict. Digital visibility does not simply standardise language; it makes the politics of standardisation more public. 136.15 Search and hashtags create pressure toward searchable spellings Search systems operate on strings, metadata and probabilistic matching. A creator who wants discovery has incentives to repeat widely recognised spellings, add alternative spellings or include Hindi and English 14031403 GAJENDRA THAKUR keywords. Hashtags similarly reward consistency because a one-character difference can fragment an audience. This may encourage practical standardisation around popular forms even without an academy or grammar committee. It can also privilege the spellings used by the largest accounts. Searchability is therefore a new source of linguistic authority—less formal than a dictionary, but powerful because it determines whether content can be found. 136.16 Recommendation systems can expand small-language reach while obscuring how selection occurs Recommendation feeds can expose users to regional-language material they did not explicitly search for. This is potentially transformative for small and dispersed linguistic publics: a song or comic clip can travel beyond subscribers and locality. Yet recommendation systems are proprietary and constantly changing. Researchers usually cannot see the full ranking logic, training data or language-identification errors. A sudden increase in visibility may reflect content quality, engagement, platform experimentation or broader trends. The historian should therefore treat algorithmic reach as a mechanism with uncertain internal evidence rather than as a transparent measure of cultural popularity. Figure 543 — Different platform affordances affect different linguistic dimensions: typing and script choice, spoken registers, audiovisual performance, searchable spelling, algorithmic amplification and AI-assisted conversion. 136.17 Monetisation can reward language mixing and scalable formats Advertising and creator-revenue systems convert attention into income, but they do not necessarily reward linguistic distinctiveness. A creator may mix Hindi with Maithili or Angika to reach a larger monetisable audience, use English words in titles for search, or produce short repeatable formats favoured by recommendation systems. This does not mean local languages disappear. It means linguistic choices are partly embedded in platform economics. The strongest cultural output may come from creators who can move between intimate local address and broader multilingual circulation. Economic incentives should therefore be analysed alongside identity and expressive preference. HISTORY OF MITHILA, VAJJI & ANGA — VOLUME II 136.18 Migration turns regional languages into translocal digital publics Migration, discussed throughout Parts XI and XIII, is one of the most important engines of online regional-language use. Migrants in Delhi, Punjab, Gujarat, Mumbai, the Gulf, Kathmandu and other destinations maintain everyday contact with home through messaging and video. Social media allows festivals, songs, local news, jokes and political debates to circulate between origin and destination in real time. The audience for a Maithili or Angika video is thus not geographically confined to its historical region. Digital language publics are translocal: they are organised by memory, kinship and interest as much as by current residence. 136.19 The India–Nepal border is porous linguistically even when digital regulation is national Madhesh and north Bihar share Maithili and other linguistic networks, but telecommunications regulation, data protection, platform law, education policy and official language regimes remain national. The same video or message can cross the open border instantly while the institutional context differs. Nepal’s 2021 census records Maithili as 41.73 percent and Bajjika as 18.44 percent of Madhesh Province’s population, alongside Bhojpuri, Nepali, Tharu, Urdu and other languages. Social media overlays this plural census geography with a cross-border communication field. Cultural connectedness should not be mistaken for administrative uniformity. 136.20 Gender and device ownership shape who speaks publicly and in which register Access to a household phone is not the same as autonomous control of a personal device. Earlier chapters showed gender gaps in personal smartphone ownership among young people even where household access was high. Shared devices can limit privacy, timing and willingness to participate in public discussion. Women may use closed messaging groups more freely than public comment sections, while safety concerns can influence profile names, photographs and language style. Social-media language data are therefore socially selected: the most visible speakers are not automatically representative of everyone who speaks the language offline. 136.21 Generational change is visible in format as much as vocabulary Younger users often move quickly among short video, memes, emoji, voice notes, gaming vocabulary, Hindi film references and English technical terms, while older users may rely more on voice calls, forwarded messages or devotional and family content. These are tendencies, not fixed age rules. The linguistic significance lies in format. A generation raised with searchable, editable and remixable media experiences language through captions, comments and clips as well as conversation and books. This expands the range of public writing while shortening many messages and increasing the role of audiovisual cues. 136.22 Memes and humour are laboratories of lexical innovation Humour rewards recognition of local accent, stereotype, proverb, kinship term and double meaning. Memes compress these into highly repeatable forms. A local word can become salient because it is attached to a recurring image or catchphrase, while code-mixing creates jokes that depend on contrast between registers. Humour can strengthen linguistic solidarity, but it can also caricature rural speech, caste, gender or district identity. The researcher should preserve both creativity and power relations: virality is not neutral evidence of acceptance. 14051405 GAJENDRA THAKUR 136.23 Music and short video create a powerful oral economy of language Songs have always travelled across Mithila, Vajji and Anga through performance, cassettes, radio, television and migration. Social platforms change the speed and granularity of circulation. A complete song can coexist with fifteen-second hooks, dance clips, devotional excerpts and user remixes. Language travels through melody even when listeners cannot read the script. This helps explain why oral and musical forms can achieve digital reach larger than prose. It also creates copyright and attribution questions when recordings are reposted without clear credit, linking the language economy back to the intellectual-property issues examined in Chapter 128. 136.24 News and public debate create a new vernacular sphere but not a single public sphere Local journalists, institutions, political actors and ordinary users now publish directly through pages, channels and groups. Regional languages can therefore enter public debate without waiting for a newspaper column or broadcast slot. Yet online publics are fragmented by platform, locality, caste/community networks, ideology and recommendation systems. A widely viewed Maithili video may reach one segment of speakers while remaining invisible to another. The existence of more vernacular content should not be equated with a unified democratic public sphere. The key change is lower publication cost combined with greater fragmentation of attention. 136.25 Misinformation travels through trusted linguistic networks as efficiently as accurate information Regional-language intimacy can increase trust. A message forwarded by a relative or recorded in a familiar accent may feel more credible than an anonymous English notice. This is valuable for public-health communication, disaster warnings and civic information, but the same trust can accelerate rumours, fabricated quotations and manipulated video. Fact-checking systems often have fewer resources for smaller languages and local dialects than for major national languages. Digital literacy therefore includes the ability to verify claims across languages and sources, not merely the ability to operate a phone. 136.26 Content moderation and automated language identification are uneven for low-resource varieties Large platforms use combinations of automated classifiers, user reports and human review. Performance is generally strongest where abundant labelled data and commercial incentives exist. Regional varieties with fluid spelling, Roman transliteration and code-mixing can be harder to classify. A system may mistake Maithili or Bajjika for Hindi, fail to recognise abusive local vocabulary, or over-remove benign expressions if context is poorly modelled. Because moderation systems are proprietary, regional error rates are rarely available. Researchers should describe this as a plausible structural risk supported by the low-resource character of many language technologies, not invent unsupported platform-specific accuracy figures. 136.27 AI translation and speech technologies create a new phase of digital language politics India’s BHASHINI programme illustrates the rapid institutional expansion of multilingual AI. The Ministry of Electronics and Information Technology’s 2025–26 annual report states that the platform supports more than 36 languages for text translation and more than 22 for voice, alongside a large multilingual glossary. Such infrastructure can improve subtitles, government services, speech interfaces and HISTORY OF MITHILA, VAJJI & ANGA — VOLUME II content discovery. Yet support counts do not guarantee equal quality across languages or varieties. Training data, dialect coverage, named entities, spelling variation and code-switching remain difficult. For Maithili, Angika and Bajjika, AI can widen access while also amplifying errors if users treat automated output as authoritative language. 136.28 Digital archives must confront ephemerality, deletion and platform dependence Social media produces enormous quantities of linguistic material but is an unstable archive. Accounts disappear, links break, platforms change policies, comments are deleted and recommendation context is lost. Screenshots preserve appearance but not searchable text or metadata; downloaded videos may lose captions and discussion threads. Long-term language history therefore requires deliberate preservation outside the platform: dated captures, transcripts, metadata, rights information and where possible original files. Digital abundance should not be confused with permanent preservation. 136.29 Measurement is the central methodological problem Official censuses measure mother tongue, not social-media language. Telecom regulators measure subscriptions and traffic, not the language of every message. IAMAI–Kantar measures national internet behaviour through surveys, while platform analytics are proprietary and creator-specific. Public search counts change over time and can be distorted by duplicate, automated or cross-linguistic content. No single dataset therefore answers “how much Maithili, Angika or Bajjika is on social media?” A rigorous account triangulates population language data, connectivity evidence, platform affordances, samples of public content and qualitative observation, and states clearly when figures are national rather than regional. 136.30 Conclusion: social media expands language domains without guaranteeing language transmission The major historical change is domain expansion. Regional languages now circulate through messaging, video, music, comments, livestreams, digital commerce and transnational family networks in ways that print- era institutions could not match. This can increase prestige, creativity and cross-border connection, especially where oral forms are strong. But digital visibility is not the same as intergenerational transmission, literacy, educational use or institutional security. A language can be highly visible in entertainment while weakening in schooling or formal writing. The future of Maithili, Angika and Bajjika will therefore depend on the interaction of homes, schools, publishing, public institutions and digital platforms rather than on social media alone. Table 136.1 — Evidence architecture for analysing social media and language Evidence source What it establishes Use in Chapter 136 Main limitation Census of India 2011 mother tongue plus baseline linguistic dated; does not C-16 / C-17 bilingualism/trilingualism repertoire and measure social-media in Bihar classification use Linguistic Survey of census-linked description shows how official classification is not India: Bihar of Bihar language classification groups identical to self- categories and names varieties identification online Nepal NPHC 2021 / mother-tongue counts cross-border census language is not Languages in Nepal and multilingual comparison; Maithili platform language 2025 structure and Bajjika in Madhesh Madhesh Province province-level language regional visualisation administrative of the multilingual province is not 14071407 GAJENDRA THAKUR Evidence source What it establishes Use in Chapter 136 Main limitation Data Portal counts from NPHC 2021 field identical to historical Mithila IAMAI–Kantar national active internet establishes scale of India-wide survey; Internet in India users and Indic-language Indic-language digital not Bihar-specific Report 2024 activities demand TRAI performance current telecom and connectivity telecom subscriptions indicators 2026 internet-system context in infrastructure do not identify India surrounding language language use Nepal broadband and mobile- connectivity context national service data Telecommunications service context on Nepal side are not Madhesh Authority MIS language-use counts Unicode Standard, encoded script repertoire technical basis for encoding does not Tirhuta block machine-readable ensure fonts, Tirhuta text keyboards or user adoption BHASHINI / MeitY multilingual AI models, institutional context support count does 2025–26 translation and voice for language not demonstrate support technology equal quality by variety Public platform actual spellings, scripts, qualitative study of non-random; samples / creator formats and audience language practice proprietary metrics; analytics interaction unstable over time Digital ethnography how users negotiate interprets social requires contextual and CMC identity, code-switching meaning rather than sampling and cannot scholarship and platform norms only counts produce a census total