This blog is part of a sunday reading Film screening activity assigned by Dr. Dilip Barad Sir following the screening of Humans in the Loop. The activity encouraged students to reflect thoughtfully on the film and engage with its ideas beyond surface-level viewing. Through this task, students were expected to observe the film’s portrayal of artificial intelligence, digital labour, and human involvement in technological systems, while also developing critical thinking and a deeper understanding of contemporary digital realities.
Pre- viewing task :
1. AI Bias & Indigenous Knowledge Systems
AI bias refers to systematic errors in machine learning outputs that reflect and reproduce the prejudices, gaps, and assumptions embedded in training data often favouring dominant cultural, racial, or geographic perspectives over marginalised ones.
Because AI models learn from human-labelled datasets, the biases of those who design, fund, and supervise the labelling process are baked directly into the technology's "worldview."
Humans in the Loop dramatises this through Nehma's work: she is asked to label ecological objects plants, insects, animals using industrial categories like "pest," "weed," or "crop" that conflict with her Oraon Adivasi knowledge system, which understands the same entities relationally and contextually.
The film raises the critical point that classifiers like "pest" or "weed" are dependent on functionality what is a weed to one party is a herb to another prompting the question: "Will this industrial consumption economy dictate our knowledge?"
Indigenous ecological knowledge (IEK) is holistic, place-based, and relational; it resists the reductive binary labelling that machine learning requires, thereby exposing how AI's very architecture privileges extractive, capitalist epistemologies.
A telling scene in the film shows an image generator producing a white boy atop an alligator when an Adivasi child asks to see himself on a crocodile a vivid illustration of whose reality gets encoded as "default."
This challenges us to ask: who defines ground truth in AI, and whose knowledge systems are rendered invisible or invalid in the process?
2. Labour & Digital Economies
Invisible labour in digital economies refers to the unacknowledged, low-paid, and often feminised human work that underpins technologies marketed as automated or artificially intelligent including data annotation, content moderation, and image labelling.
The term "artificial intelligence" itself obscures the vast human infrastructure behind it: over 70,000 Indians mostly rural women form AI's invisible workforce, performing the repetitive cognitive tasks that train global AI systems.
This labour is "invisible" in multiple senses: it is geographically remote, socially marginalised, algorithmically erased from the final product, and absent from mainstream narratives about technological progress.
The film is attentive to the texture of this labour fluorescent-lit rooms, lagging computers, supervisors chasing targets, and workers trying to make sense of labels written in a different idiom from their own lives.
Highlighting such labour matters because it dismantles the myth of AI as disembodied and neutral; it reveals supply chains of exploitation that mirror older colonial and caste-based patterns of resource extraction.
It also raises questions of fair compensation, intellectual credit, and whose cognitive labour is commodified without consent or recognition in the building of billion-dollar AI industries.
The film highlights the unsung human labour that quietly powers the global artificial intelligence industry, insisting this story be central, not peripheral, to how we narrate technological modernity.
3. Politics of Representation
Representation in Humans in the Loop operates on two interlocking levels: how AI systems represent (or fail to represent) Adivasi people, and how the film itself represents both technology and Adivasi culture to mainstream audiences.
Critics and publicity materials emphasise the film's unusual vantage point it examines how technological progress can entrench exclusion while sidelining indigenous knowledge systems positioning Adivasi experience not as a backdrop but as the analytical lens through which AI is scrutinised.
The film's most cited scene where an AI image generator produces a stereotyped, Europeanised image when asked to depict a tribal woman directly indicts the representational politics of training data, showing that AI does not see Adivasi people at all, or sees them through a colonial distortion.
Executive producer Kiran Rao framed the film's intervention explicitly around "equitability, representation and data colonialism," signalling that representation is understood here as a political and structural issue, not merely an aesthetic one.
On the question of how Adivasi culture is depicted, reviews are divided: some praise the film's grounded, research-based authenticity, while at least one Letterboxd critic accused it of "fetishising" Adivasi life for a liberal festival audience without genuine political commitment.
The film's bilingual Hindi-Kurukh language choice is itself a representational act giving screen presence to an Adivasi language rarely heard in mainstream Indian cinema.
Taken together, the film asks audiences to be conscious of the double bind: AI misrepresents Adivasi communities, and cinema, if not careful, can do the same. While- watching task: 1.NARRATIVE & STORYTELLING
How does the film situate Nehma’s personal life with larger algorithmic structures? What narrative turns foreground labour, family, and knowledge systems?When Nehma “teaches” AI, what does this suggest about human-machine learning loops beyond technological jargon?
In Humans in the Loop, Nehma’s act of “teaching” AI through data annotation reveals that machine learning is not an autonomous or purely technical process but a deeply human-dependent cycle. Her work shows that AI systems learn by absorbing human judgement, interpretation, and cultural assumptions. In this sense, the human–machine learning loop becomes a process of knowledge transfer where people quietly shape the intelligence that machines appear to possess.
Beyond technological jargon, Nehma’s teaching suggests that AI learning is also emotional, ethical, and interpretative labour. Each label she assigns reflects decisions about meaning, context, and relevance, demonstrating that machines do not understand the world directly but through human mediation. This reframes AI as a collaborative yet unequal relationship: while humans train and correct the system, their contributions often remain invisible and undervalued.
Moreover, the film highlights that what Nehma teaches AI is influenced by her own worldview, even when constrained by industrial categories. This indicates that human–machine learning loops are also epistemological loops, where certain knowledge systems are translated into machine-readable forms while others are simplified or excluded. Thus, Nehma’s role suggests that AI is not merely learning facts but inheriting human perspectives, biases, and limitations.
Ultimately, the film invites us to see the human–machine loop as a socio-cultural process rather than a purely technical one, raising critical questions about authorship, agency, and recognition: if humans teach AI, whose knowledge is being amplified, and whose remains unheard?
2. REPRESENTATION & CULTURAL CONTEXT
How are Adivasi culture, language, tradition, and ecological knowledge represented?
In Humans in the Loop, Adivasi culture, language, tradition, and ecological knowledge are represented with intimacy and everyday realism rather than through romanticised or exotic portrayals. The film situates these elements within Nehma’s daily life, showing how her identity as an Oraon Adivasi woman shapes the way she understands nature, work, and community.
Culture and tradition are reflected through domestic spaces, family interactions, and community practices that emphasise collective living and continuity with ancestral ways of life. The film presents rituals, food habits, and local customs subtly in the background, allowing viewers to see Adivasi life as lived experience rather than spectacle. This grounded representation highlights the coexistence of tradition and modern digital labour.
Language plays an important role in conveying identity and worldview. Nehma’s use of her local language in conversations contrasts with the English-dominated digital interface of annotation platforms, illustrating a linguistic hierarchy embedded within global technological systems. This tension suggests that while Adivasi language carries cultural memory and meaning, it remains marginal within digital infrastructures.
Ecological knowledge is one of the film’s most powerful representations. Nehma’s understanding of plants, insects, and animals is relational and context-based, shaped by lived interaction with the environment. The film contrasts this holistic perspective with the reductive labels required by AI annotation, such as “weed” or “pest,” thereby revealing how indigenous ecological knowledge challenges industrial and algorithmic framings of nature.
Overall, the film portrays Adivasi identity as dynamic and resilient, showing how cultural memory, language, and ecological wisdom persist even within globalised technological networks. At the same time, it raises critical questions about visibility and recognition, suggesting that while Adivasi knowledge sustains both community life and digital economies, it often remains under-acknowledged in dominant narratives of technology and progress.
Does the film challenge or reinforce dominant media stereotypes about tribal communities and modern technology?
In Humans in the Loop, the portrayal of Nehma and her community largely challenges dominant media stereotypes about tribal (Adivasi) communities and their relationship with modern technology.
Mainstream media often frames tribal communities as either primitive, technologically disconnected, or romantically close to nature but outside modernity. The film disrupts this binary by presenting Nehma as both deeply rooted in her Adivasi culture and actively participating in global digital labour. Her role as a data annotator demonstrates that tribal communities are not excluded from technological systems; instead, they are integral yet often invisible contributors to AI development.
The film also challenges stereotypes by foregrounding agency and intellectual labour. Rather than depicting Nehma as a passive subject of development narratives, it shows her making interpretative decisions, negotiating meaning, and effectively “teaching” AI. This representation counters the idea that technological knowledge belongs only to urban or elite populations.
At the same time, the film does not ignore structural inequalities. By showing the precarious nature of annotation work and the epistemic tensions between indigenous ecological knowledge and industrial AI categories, it reveals how dominant technological systems can marginalise local knowledge. This nuanced portrayal avoids reinforcing stereotypes and instead exposes the power dynamics that shape visibility, labour value, and knowledge recognition.
Ultimately, the film presents Adivasi identity as modern, adaptive, and intellectually engaged, challenging simplistic representations and encouraging viewers to rethink who participates in technological futures. It suggests that the real issue is not technological absence but unequal recognition and representation within digital economies.
3. CINEMATIC STYLE & MEANING
- Wide-angle shots embed characters within nature they are not against the landscape but part of it.
- Natural, dappled light warm, diffuse, non-directional. Nothing is spotlit; everything participates equally.
- Low, ground-level camera in the porcupine sequence places human and animal at the same horizontal plane, encoding a relational, non-hierarchical world-view.
- Compositions are deliberately non-geometric: roots, grass, and canopy break the frame into irregular, organic shapes the visual opposite of the grid.
- Tight mid-shots and close-ups compress space walls, screens, and ceilings are always visible, implying no world beyond the frame.
- Fluorescent artificial light flat, shadowless, institutional. It erases depth and complexity, visually rhyming with the binary label: everything is either lit or not.
- Cool blue-grey colour palette contrasts directly with the warm ochres of the forest sequences.
- The computer screen as light source it illuminates Nehma's face, not the other way around. The machine directs light onto the human; power flows from screen to body.
- Workers blocked in rows, facing screens — assembly-line composition that places data labour within the visual grammar of industrial production.
- Medium close-ups that are tender, not confining the camera holds rather than traps.
- Tactile texture rock, bark, weave, soil emphasised through close framing. A visual argument that Adivasi knowledge is embodied and material, not digitisable.
- Horizontal eye-lines between Nehma and her children in knowledge-transmission scenes equal, not hierarchical contrasting directly with the top-down blocking of supervisors over workers in the data centre.
The Sound Team
Sound Design: Kalhan Raina
Score: Saransh "Khwabgah" Sharma
Editing: Swaroop Reghu & Aranya Sahay
These three collaborators divide their responsibilities clearly: Raina governs the diegetic world (what exists within the film's reality), Sharma governs the non-diegetic emotional register (the score we hear as audience), and the editors govern rhythm and time how quickly or slowly the film breathes.
Sound Design: Diegetic Contrast
The most fundamental sonic opposition in the film is between organic ambient sound and mechanical ambient sound.
In the forest and domestic world:
Birdsong, wind through grass, the sounds of water, animal movement, children's voices, communal singing all are irregular, unpredictable, and layered simultaneously. No single sound dominates. The soundscape is polyphonic and relational, mirroring the Adivasi ecological world-view the film endorses.
Crucially, these sounds are not cleaned up in post-production. The presence of background noise, ambient texture, and sonic imperfection is a deliberate choice it signals authenticity, embodiment, and the irreducibility of living sound to pure signal.
In the data centre:
The sonic world contracts to repetitive mechanical sounds: keyboard clicks, mouse clicks, the hum of fluorescent tubes, the lag and boot sounds of slow computers. These sounds are metronomic they have regularity, uniformity, and no organic variation.
The click of the mouse labelling an image becomes the film's most politically loaded sound. Each click is an act of classification pest or not pest, weed or crop, human or other. The sound is small, dry, and final. It enacts the violence of binary categorisation in sonic form.
The Score: Sharma's Compositional Strategy
Sharma's music blends organic soundscapes with textured electronic production drawing from ambient, post-classical, and downtempo influences. Guitar, piano, synthesizers, and field recordings form the basis of his pieces, built on layers that feel intimate, tactile, and reflective. For Humans in the Loop, the aim was to create a subtle bed of sound that would allow the emotion to unfold naturally.
This compositional strategy is not merely aesthetic it is politically calibrated:
The score uses both organic instruments (guitar, piano) and synthetic textures (synthesizers, electronics) simultaneously. It refuses to assign "natural instruments" only to forest scenes and "electronic sounds" only to data centre scenes. Instead, it holds both registers in one sonic space the same way Nehma holds both worlds in her body.
\
Sharma's ambient soundtrack infuses a dream-like quality The Federal particularly in the opening porcupine sequence here the score functions as sonic mise-en-scène, enlarging a small, quiet visual event into something of mythic weight without dramatising it.
The score does not swell or signal. It operates in understatement. This restraint prevents the audience from being told how to feel, placing the interpretive burden back on the viewer appropriate for a film about who has the right to assign meaning.
The Kurukh Music Problem: Authenticity vs. Accessibility
This is the film's most sophisticated and honest sound decision. Director Sahay explains:
Oraon music is very different from the music that the broader Indian audience listens to. It does not follow the four-by-two meter. Musicians often change the key, sometimes even in the middle of a song and may also change the rhythm in the middle of a song. It was difficult to decide how to bridge the music: to not alienate the audience completely, but to also retain authenticity. This is where the synth, the violin, and other instruments came in.
This is a representational dilemma made audible. The film's score is itself a negotiation between two knowledge systems the irregular, non-metrical rhythms of Kurukh Adivasi music and the regularised, metrically predictable structures of mainstream Indian and global cinema audiences' expectations. The synthetic instruments act as a bridge familiar enough to hold the audience, unfamiliar enough to signal that they are in a different sonic world. Importantly, the director acknowledges the compromise honestly, which itself is an ethical act.
Editing Rhythms: Pacing as Political Argument
The film is edited by Swaroop Reghu and Sahay himself a significant choice, as the director controlling the edit ensures that pacing decisions are inseparable from thematic ones.
Slow rhythm in analog sequences:
Forest and domestic scenes are given longer takes and fewer cuts. Time is allowed to accumulate. A scene of Nehma and the porcupine, or Nehma pointing out living things to her children, breathes the edit does not hurry it. This slow rhythm encodes a different relationship with time: non-industrial, non-target-driven, attentive.
Tighter rhythm in labour sequences:
Inside the data centre, cuts are more frequent. The editing rhythm begins to mirror the click-rate of the labelling work image after image, classification after classification. The edit trains you, as a viewer, to process images quickly, just as Nehma is trained to. This is a formally self-aware gesture: the editing makes you complicit in the very cognitive rhythm it is critiquing.
The central parallel edit AI infant / Guntu:
The film's most powerful editorial sequence cuts between Nehma labelling infant muscle-movement data at her workstation and close-ups of her own son Guntu's limbs at home. The editing holds both in the same rhythm the cuts between them are not accelerated or slowed. By keeping the temporal rhythm identical, the edit argues that these two events are not merely analogous but materially linked: the same unit of time, the same unit of attention, the same body displaced from one to the other. No score is needed here. The argument is made entirely through the cut.
Restraint and silence:
Sahay exercises admirable restraint with a focus on silences and moments of introspective quiet that allows the film's themes to breathe without descending into didacticism. The Federal Silence in this film is not absence it is where the film places its most serious claims. When Nehma refuses to label the caterpillar as a pest and is reprimanded, the scene is followed by a held silence rather than a dramatic score cue. The silence asks: who fills this space? Who has the right to speak after this? The audience is left in the gap.
4. ETHICAL & POLITICAL QUESTIONS
1. Whose Categories Govern? The Epistemological Dilemma
Nehma gets in trouble for failing to label a creature as a pest, because based on her community's ecological knowledge, she knows it is not harmful to crops. The film uses this scenario to highlight how AI relies on universal categories, often ignoring valuable local and indigenous knowledge when the world is forced into a handful of machine-readable boxes, nuance is the first casualty.
The ethical dilemma here is foundational: who has the authority to define "correct"? The AI system's categories are designed for an extractive agricultural economy by clients in the Global North. Nehma's knowledge built from generational, place-based observation is more accurate but inadmissible. The system treats superior knowledge as error.
2. Extraction Without Compensation The Data Colonialism Dilemma
Representative datasets, containing images of indigenous people, their languages, culture and knowledge systems, are a product of the labour of the masses. The AI image generator is arguably better off with Nehma's additions but what does she or her community get out of it? If foreign AI companies are dependent on the value generated by India's tribals, trained on their images and utilising their knowledge of nature, what is their return?
Nehma's cultural knowledge and her community's images are incorporated into a commercial AI system she will never own, profit from, or have any control over. This is data colonialism the extraction of epistemic and representational value from a marginalised community without consent, credit, or compensation.
3. Bias Reproduction The Inheritance Dilemma
In a world where machines learn to absorb human biases, Nehma comes to understand that the technology she has undertaken, like tending her children, inherits the discrimination of its labeller she can't help but question whether the system she creates will only serve to perpetuate the fate she has suffered.
The AI does not generate neutral outputs. It inherits the biases of whoever designed its classification framework. AI rides on the back of struggling low-income women labourers, and their input or their lives is rarely acknowledged a biased AI leads to misrepresentation and further isolates already marginalised communities. The ethical dilemma: Nehma is both the victim of existing bias and, through her constrained labour, the unwilling reproducer of new bias.
4. Representational Violence The Image Generation Dilemma
When Nehma feeds her own image and those of her community into the AI generator to correct its misrepresentations, the film raises a fresh dilemma: culturally representative datasets are something AI companies are increasingly hungry for fresh concerns arise that do not get addressed in the movie. These representative datasets are a product of the labour of the masses, but the communities supplying them receive nothing in return.
Correcting the AI's blindness to Adivasi faces requires Nehma to submit her own image to the same system that previously erased her. To be seen, she must surrender her image to an economy that does not recognise her humanity. This is the film's most precise ethical trap.
5. The Hierarchy of Accountability The Structural Dilemma
The film shows how dataset guidelines, client briefs, and workplace hierarchies encode dominant assumptions into the material Nehma is asked to produce tracing how AI's "neutrality" is undermined by who gets to name things, whose faces are over-represented, and whose stories never enter the frame.
The ethical chain runs: American tech client → Indian centre manager → Adivasi data worker. Responsibility is diffused across this chain so that no single actor appears culpable. Director Sahay asks: "When are we going to take responsibility as humanity for the kind of algorithms we're building?" The film's answer is that the architecture of outsourcing is designed to prevent that accountability from settling anywhere. Core Ethical Argument
All five dilemmas converge on a single point: the ethics of AI cannot be separated from questions of caste, gender, and geography. The film insists that training data is never culturally neutral it always encodes the power relations of those who design, fund, and supervise its collection. When that data is sourced from communities with the least power in the global economy, the ethical debt compounds at every layer of the system.
The "Human-in-the-Loop" Metaphor Beyond the Technical Term
The Technical Term Briefly
In machine learning, human-in-the-loop (HITL) refers to a system design where a human overseer corrects algorithmic errors, validates outputs, and improves the model's accuracy over iterative cycles. The human is positioned as a check on the machine. The term implies agency, authority, and mutual benefit.
The film accepts this definition as its starting point then systematically dismantles every assumption embedded in it.
Politically: The Loop as Colonial Circuit
The title alludes to the closed-loop relationship between humans and technology one "programs" the other and vice versa, forever. But the film makes visible that this loop is not symmetrical. It has a direction: value flows upward from Jharkhand to Silicon Valley; constraint flows downward in return.
The political loop the film exposes is data colonialism a 21st century structure that mirrors 19th century colonial extraction. Raw material (Nehma's cognitive labour and cultural knowledge) is extracted from the periphery, processed at the centre, and returned as a finished product (AI) in which the original producer has no ownership, profit, or control. The film engages one of the most urgent global conversations of our time: how we can prevent social inequities from making their way into AI and instead use it to enhance underrepresented voices.
Crucially, the political loop is self-reinforcing. The AI trained on biased data produces biased outputs. Those biased outputs further marginalise the communities whose knowledge was extracted to build the system. Nehma corrects the machine, but the machine's economic architecture is correcting her in return telling her which knowledge is valid, which categories are acceptable, which version of the world is real. She is in the loop, but the loop is not in her interest.
Socially: The Loop as Caste and Gender Structure
The film uses Nehma's work at the AI data lab as a site for exploring themes of caste, assimilation, and the desire to belong to a world that seems more "legitimate" in the eyes of dominant society.
Nehma enters the film already caught in multiple social loops she did not design:
The loop of gender: as a woman, her labour domestic and digital is treated as naturally available, endlessly renewable, and minimally compensable. There may even be a hidden metaphor for the way our world operates characterised by a lack of more feminine and nurturing values in favour of more masculine ambitions of dominance and control. The loop of caste: a profession that thrives on binary labels is outsourced to people whose plurality transcends labels. Humans are instructed to think like machines in order to instruct machines to act like humans. A marginalised Indian's social conditioning is at odds with a job that formalises societal bias.
The loop of class: Dhaanu, Nehma's daughter, prefers her urban, upper-caste father gravitating toward a world that appears more "legitimate." This is the social loop working through the next generation: assimilation reproducing the very hierarchy that excludes the assimilated.
The loop metaphor here means that social exclusion is not a one-time event but a recursive system one that reproduces itself across generations, across institutions, and now across algorithms.
Culturally: The Loop as Epistemological Erasure
This is the film's deepest and most original extension of the metaphor. In a training session at the AI centre, Nehma's supervisor tells her that artificial intelligence is like a child it must be taught how to see the world. This metaphor becomes a central ideological battleground. If AI is indeed a child, who becomes the teacher? What values and assumptions are encoded in the data it consumes? The cultural loop operates as follows: Nehma possesses a sophisticated, relational, ecologically precise knowledge system. She is hired to transmit this and other knowledge to an AI. But the transmission is filtered through categories she did not design. A caterpillar must be either a pest or not a pest. A plant must be a weed or a crop. The cultural loop does not absorb Nehma's knowledge it translates it into a form that erases what makes it valuable, then feeds it back into the world as "objective data." The AI learns from her but cannot see her.
This is the cultural loop in its most precise form: unlike the thousands of images and videos she goes through every day, Nehma isn't even afforded the dignity of being a data point. She teaches the system to see, but the system cannot return the gaze.
The Reciprocal Loop: Who Is Training Whom?
The film's sharpest political-cultural insight is that the loop runs in both directions but not equally. One "programs" the other and vice versa, forever. The machine trains on Nehma's knowledge. But Nehma is simultaneously being trained by the machine being told to suppress her ecological intuition, to accept industrial categories, to think in binaries. To programme the robot, she must think robotically too.
The cultural violence of the loop is therefore not just extraction it is substitution: the AI economy does not merely take Nehma's knowledge; it replaces it with a degraded, commodified version of itself and asks her to use that replacement as her new cognitive standard.
A Dissenting Voice: What the Loop May Not Resolve
It is worth noting that not all critics accept the film's political architecture as fully realised. The film claims to speak about marginalisation while marginalising the very community it depicts. It critiques data bias but exhibits cultural bias. It condemns invisibility but erases the Adivasi lived experience. This critique from within the Adivasi community points to an uncomfortable meta-irony: the film itself may be operating in a loop analogous to the one it critiques, extracting Adivasi cultural material for a global festival circuit audience without sufficient return to the community depicted.
This does not invalidate the metaphor. It deepens it suggesting that the human-in-the-loop problem is not confined to the data centre but extends to the film industry, the festival circuit, and the very act of representation itself.
No comments:
Post a Comment