M.A. (Women's Studies) GENDER AND EMERGING TECHNOLOGY
Notes Prepared By
UNIT 1: ALGORITHMS, AI AND CODED BIAS
1. Understanding How “Neutral” Algorithms Inherit Human Prejudices
1.1 What is an Algorithm?
An algorithm is a structured procedure or set of rules used to process information and produce an output.
For example:
Input → Processing/Rules → Output
A recruitment system may receive:
CV + educational qualifications + experience + other information
and produce:
Shortlisted / Not shortlisted
An algorithm itself does not necessarily have intentions or prejudices. However, the data, objectives, categories, assumptions and human decisions incorporated into an algorithm can reproduce existing social inequalities.
NIST specifically emphasizes that AI bias is not only a technical problem. It can originate from datasets, algorithmic processes, human decision-making, institutions and wider society. (NIST Publications)
1.2 Why Are Algorithms Not Always Neutral?
A common assumption is:
Human beings are biased → computers are objective.
But this is misleading.
A better understanding is:
Human society → data → algorithm → decision → social consequences
If the society producing the data is unequal, the data may contain those inequalities.
Example
Suppose a company has historically hired mostly men for senior technical positions.
The company's historical recruitment data therefore contains:
Male applicants → more previous selections → apparently “successful” employees
If an AI system is trained to identify characteristics associated with previous successful employees, it may learn that characteristics statistically associated with male applicants are desirable.
The algorithm does not have to “hate women.”
Instead:
Historical inequality → biased training data → algorithmic pattern → unequal outcome
NIST explains that statistical/computational bias can emerge from non-representative data, while human and systemic biases can enter throughout the AI lifecycle. (NIST Publications)
1.3 Types of AI Bias
A. Historical/Systemic Bias
This occurs when existing social inequalities become embedded in data or institutional practices.
Example:
If women historically received fewer promotions, historical promotion data may reflect that inequality.
NIST defines systemic bias as bias resulting from rules, processes or norms that advantage some social groups while disadvantaging others. (NIST Computer Security Resource Center)
B. Data/Representation Bias
A dataset may not adequately represent the population on which the technology will be used.
Example:
If a facial-recognition system is trained or tested disproportionately on certain demographic groups, its performance may differ across groups.
NIST recommends checking whether datasets adequately represent the populations in the context in which an AI system will actually operate. It also recommends disaggregated testing by characteristics such as gender, age, ethnicity, ability and region. (NIST SI Resource Center)
C. Measurement Bias
The system may measure an imperfect proxy for what we actually want to know.
For example:
“Employment suitability” cannot be directly observed.
A system might instead use:
educational institution,
postcode,
previous salary,
employment history,
social-media behaviour.
But these variables can themselves reflect social inequality.
NIST notes that proxies can produce harmful or discriminatory outcomes when they imperfectly represent the concept being measured. (NIST)
D. Human-Cognitive Bias
Humans designing or using AI also make judgments.
For example, a manager may think:
“The AI recommended this candidate, so the recommendation must be objective.”
This creates automation bias—excessive trust in an automated output.
NIST identifies human-cognitive bias as one of the major categories of AI bias. (NIST Publications)
1.4 Facial Recognition and Gender/Racial Bias
Facial recognition provides an important example of why supposedly “neutral” technology must be critically examined.
NIST's Face Recognition Vendor Test (FRVT) Part 3: Demographic Effects examined facial-recognition algorithms across demographic groups.
It found demographic differences in performance across many algorithms. In particular, false-positive rates varied substantially across demographic groups, and false positives were generally higher for women than men in the datasets examined. (NIST)
What is a false positive?
A false positive occurs when the system incorrectly identifies two different people as the same person.
Why is this important?
Imagine:
Person A → Facial recognition system → Incorrectly matched with Person B
In a low-risk situation, this may simply be inconvenient.
But in:
policing,
border control,
security,
surveillance,
an incorrect match may have serious consequences.
NIST therefore stresses that the consequences of demographic differences depend on the specific task and context of use. (NIST)
Gender perspective
This shows that:
Technology does not operate outside society.
Questions of gender, race, age and power can become questions of technological accuracy.
1.5 AI and Hiring Software
AI is increasingly used in employment processes, including:
résumé screening,
candidate ranking,
video-interview analysis,
job advertising,
workplace monitoring,
promotion,
pay decisions,
termination decisions.
The U.S. Equal Employment Opportunity Commission explicitly recognizes these as areas in which AI and automated technologies may influence employment decisions. (EEOC)
How can gender bias enter?
Suppose historical employees were predominantly men.
The system may learn patterns associated with:
previous hiring → perceived “success”
and reproduce them.
Another problem is the use of seemingly neutral characteristics that indirectly correlate with protected characteristics.
Important principle
A system does not become fair simply because a human has replaced a human decision-maker with a computer.
The EEOC emphasizes that existing employment-discrimination protections continue to apply when employers use AI. (EEOC)
1.6 Feminist Analysis of Algorithmic Bias
A feminist approach asks:
Who designed the system?
Whose data was collected?
Whose experiences are missing?
Who benefits from the system?
Who bears the risks?
Who gets to challenge the decision?
Are gender, caste, class, race, disability and location considered together?
This is important because gender discrimination rarely exists alone.
For example:
Woman + rural + poor + disabled
may experience technology very differently from:
Urban + wealthy + highly educated woman.
Thus, intersectionality is essential to the study of technological inequality.
Feminist Analysis of Algorithmic Bias: A Detailed Explanation
1. Introduction: Why should feminists study algorithms?
An algorithm is a set of computational rules used to process information and produce an output. Artificial intelligence (AI) systems may use algorithms to screen job applications, recognize faces, assess creditworthiness, recommend content, or support decisions in healthcare.
At first glance, these systems may appear neutral because they use mathematical calculations rather than personal opinions. However, mathematical processing does not automatically guarantee fair outcomes. Algorithms are developed by people, trained or configured using particular data, and deployed within social institutions that already have rules, assumptions and inequalities.
The National Institute of Standards and Technology (NIST) explains that AI bias can arise from data, human decisions, technical processes and wider systemic conditions—not just from programming errors.
Feminist analysis therefore moves beyond the question, “Does the algorithm work accurately?” It also asks:
Does it work equally well for different groups?
Whose experiences have shaped its design?
Does it reinforce existing gender inequalities?
Who gains opportunities, and who is excluded?
Can people challenge decisions that affect their lives?
Central argument: Algorithmic bias is not merely a technical problem. It can also be a problem of social power, representation, discrimination and justice.
2. Understanding the relationship between algorithms and social inequality
Consider the following process.
Existing social inequalities
Gender discrimination, caste hierarchy, class inequality and unequal access to education
Data and design choices
Whose information is collected, which variables are selected and what the system is designed to optimize
Algorithmic decision
A person is ranked, classified, recommended or denied an opportunity
Possible social consequences
Unequal access to employment, services, recognition or resources
This is a possible pathway, not an inevitable outcome of every algorithm. The important point is that social inequalities can enter a technical system at several stages and may be reproduced if they are not identified and addressed. This socio-technical perspective is discussed in NIST's Towards a Standard for Identifying and Managing Bias in Artificial Intelligence and in research on algorithmic discrimination in India.
Let's examine the seven questions in detail.
3. Who designed the system?
Meaning
This question examines the people, organizations and institutions that make decisions about how an AI system should operate.
Designers may decide:
What problem the technology should solve.
What counts as a successful outcome.
Which information the system should use.
Which groups should be represented in testing.
What kinds of errors are considered acceptable.
Whether the system prioritizes efficiency, profit, accuracy or fairness.
These decisions are not always purely technical. They can reflect organizational priorities and assumptions about society. NIST's framework treats AI risks as arising across the system's lifecycle, including design, development, deployment and evaluation.
Example: An AI recruitment system
Imagine a company developing software to shortlist candidates for managerial positions. The developers choose to train the software on the company's previous hiring and performance records.
If previous recruitment practices systematically favoured men, the historical records may contain fewer women in senior positions. A system that learns patterns from those records could reproduce some of the same disadvantages.
The system does not need to be explicitly programmed to reject women for this to happen. The risk comes from the combination of historical data, design choices and the way the model is evaluated.
This example illustrates a well-known concern in research on algorithmic hiring and gender discrimination. It should be understood as a possible mechanism rather than a claim that every recruitment algorithm discriminates.
Feminist interpretation
Feminist analysis asks whether the people designing a system understand the experiences of the people affected by it.
A team with limited experience of rural women's lives, disability, caste discrimination or unequal access to technology may overlook relevant barriers. Diversity alone cannot guarantee fairness, but inclusive participation can help identify assumptions that a more homogeneous team might miss.
Key lesson: Who designs a system matters because design decisions influence whose needs the system recognizes and whose problems it overlooks.
4. Whose data was collected?
Meaning
AI systems learn patterns from data or use data supplied to them. Data may include photographs, employment records, medical histories, educational qualifications, income, location, language and online behaviour.
A feminist approach asks whether the data adequately represents the people who will be affected by the system.
This issue is central to Data Feminism by Catherine D'Ignazio and Lauren F. Klein, which argues that data collection and analysis are connected to power and that data must be understood within its social context.
Example: Facial recognition
In their 2018 study, “Gender Shades: Intersectional Accuracy Disparities in Commercial Gender Classification,” Joy Buolamwini and Timnit Gebru evaluated three commercial gender-classification systems.
Their research found that:
The benchmark datasets they examined were disproportionately composed of lighter-skinned people.
All three evaluated systems performed worse on female faces than on male faces.
All three performed worst on darker-skinned female faces.
Error rates for darker-skinned female faces reached 34.7%, compared with a maximum of 0.8% for lighter-skinned male faces in the reported evaluation.
These figures relate to the specific datasets and systems tested in that study; they should not be generalized to every current facial-analysis system.
What does this tell us?
Suppose a facial-analysis system is tested mainly on lighter-skinned faces. Developers may conclude that it performs well overall. But its average accuracy can hide much poorer performance for darker-skinned women.
This raises three questions:
Who was included in the dataset?
Who was underrepresented or absent?
Was the system tested separately on different demographic groups?
Feminist interpretation
Data gaps can become equality gaps when missing or poorly represented groups experience less reliable automated decisions.
However, collecting more data is not always sufficient. Data must also be relevant, appropriately collected, accurately labelled and handled with respect for privacy. NIST identifies dataset quality, evaluation practices and human factors as important areas for managing AI bias.
Key lesson: Before asking whether an algorithm is fair, ask whether the data on which it relies adequately represents the people whose lives it may affect.
5. Who bears the risks?
5.1 Meaning and explanation
When an AI system is introduced, its benefits and risks may not be distributed equally among the people affected by it.
For example, a company might benefit from an automated recruitment system because it can process thousands of applications quickly. However, applicants may face unfair rejection if the system relies on biased historical data or unsuitable selection criteria.
The National Institute of Standards and Technology (NIST) explains that AI bias can harm individuals and communities even when discrimination is not intentional. It also emphasizes that bias can originate in technical processes, human decisions and wider social structures.
A feminist approach therefore asks:
Who is most likely to experience an incorrect decision?
Which groups are more exposed to harm?
Who has the resources to protect themselves?
Who is held responsible when the system causes harm?
Do affected people have access to remedies?
5.2 Example: Facial recognition and darker-skinned women
Add to Favorites
In their 2018 study, Gender Shades: Intersectional Accuracy Disparities in Commercial Gender Classification, Joy Buolamwini and Timnit Gebru evaluated three commercial gender-classification systems.
They found that darker-skinned women experienced the highest error rates among the evaluated groups. The maximum error rate for darker-skinned female faces was 34.7%, compared with a maximum of 0.8% for lighter-skinned male faces in the study. These results apply to the specific systems and benchmark dataset examined, not to every facial-analysis system today.
Why is this important?
Suppose an automated facial-analysis system is used in an important institutional process. If its performance is less reliable for a particular group, people in that group may face a greater risk of incorrect outcomes.
The seriousness of that risk depends on the application. An error in a photo-filter application is different from an error in a high-stakes identity-verification or public-service system.
The research establishes differences in system performance; the consequences must be assessed in the actual context of use. NIST similarly recommends examining both the performance of AI systems and their potential impacts on affected people.
5.3 Example: Automated decisions in public services
Imagine that a public-service agency uses an algorithm to prioritize applications for assistance.
If the system relies heavily on digital records, people with incomplete records or limited access to online services might be assessed less favourably. This could create barriers for some disadvantaged applicants.
This is a hypothetical example, not a claim that a particular government system produces this outcome.
NIST recommends considering the people and communities affected by an AI system, including how differences in data representation and digital access may contribute to risk.
5.4 Feminist interpretation
The central issue is the unequal distribution of risk.
A system may offer substantial benefits to an organization while placing a disproportionate burden on particular groups. Feminist analysis asks whether those groups have been considered in the design and evaluation process and whether their interests receive adequate protection.
NIST's framework emphasizes that fairness involves more than balanced demographic statistics: a system may still be unfair if it is inaccessible to people with disabilities or reinforces existing social inequalities.
Key takeaway: A feminist evaluation of AI must ask not only whether the system is efficient, but also who is most likely to suffer when it makes a mistake.
Point 6. Who gets to challenge the decision?
6.1 Meaning and explanation
When an algorithm makes or influences a decision about a person's life, that person should have a meaningful opportunity to question an incorrect or unfair outcome.
For example, AI systems may influence decisions involving:
Recruitment and promotion.
Loan applications.
Access to public services.
Insurance eligibility.
Content moderation and account suspension.
A person may not know that an algorithm influenced the decision. Even if they know, they may not understand why the decision was made or how to challenge it.
NIST's AI Risk Management Framework identifies accountability, transparency, explainability and interpretability as important characteristics of trustworthy AI. It distinguishes transparency—information about what happened—from explainability, which concerns how a system operates, and interpretability, which concerns the meaning of its outputs in context.
6.2 Example: A woman rejected by an AI recruitment system
Imagine a qualified woman applying for a university teaching position. An automated recruitment system ranks her application below those of other applicants.
She wants to know:
Was her application assessed correctly?
Were her qualifications recognized?
Did the system treat a career break as a negative factor?
Was the system tested for gender-related disparities?
Can a human reviewer reconsider the result?
Suppose the institution provides no meaningful explanation or appeal mechanism. The applicant may have difficulty identifying whether the decision resulted from a genuine qualification gap, incorrect data or a problematic automated assessment.
This is a hypothetical example illustrating the importance of accountability, not evidence that all recruitment systems operate this way.
Research by Inioluwa Deborah Raji and colleagues proposes an end-to-end framework for internal algorithmic auditing to help organizations identify and document potential harms throughout AI development and deployment. Their research also highlights the difficulty of tracing problems back to their source after a system has been deployed.
6.3 Why is this a feminist issue?
Feminist theory examines relationships of power. When an institution uses an algorithm to rank, classify or reject individuals, it exercises power over their opportunities.
If the affected person cannot question the outcome, the institution's decision-making power may become difficult to challenge.
This concern becomes particularly important for people who already face barriers to representation, legal assistance, education or institutional participation.
Research on algorithmic fairness in India by Nithya Sambasivan and colleagues emphasizes that the distance between AI developers and marginalized communities can make it difficult for affected people to influence how systems are designed and evaluated. Their findings are based on 36 qualitative interviews and a discourse analysis of AI deployments in India.
6.4 What would a fairer system look like?
A responsible institution could establish the following safeguards:
| Safeguard | Purpose |
|---|---|
| Clear notification | Inform people when AI significantly influences a decision. |
| Appropriate explanation | Help people understand the relevant reasons for an outcome. |
| Data correction | Allow incorrect personal information to be corrected. |
| Human review | Provide meaningful reconsideration of consequential or disputed decisions. |
| Complaint mechanism | Give affected people an accessible route to raise concerns. |
| Independent audit | Examine whether the system produces unjustified disparities. |
These safeguards are consistent with NIST's emphasis on accountability, transparency and risk management, and with Raji and colleagues' work on algorithmic auditing. The exact legal rights and obligations depend on the jurisdiction and the application.
Feminist interpretation: Fairness is not simply about receiving a decision that appears reasonable. It also concerns whether people have the power to question decisions that affect them.
Key takeaway: A system that cannot be meaningfully challenged can reinforce institutional inequalities, particularly when affected people already have limited power.
Point 7. Are gender, caste, class, race, disability and location considered together?
7.1 Understanding intersectionality
Intersectionality is a framework associated particularly with legal scholar Kimberlé Crenshaw. In her 1989 article, “Demarginalizing the Intersection of Race and Sex,” Crenshaw demonstrated how single-axis approaches to discrimination could overlook the distinctive experiences of Black women.
Her central insight was that discrimination cannot always be adequately understood by examining gender or race independently. The interaction of social structures can produce experiences that neither category captures on its own.
Applied to AI, intersectionality asks whether a system affects different groups of women differently according to other relevant aspects of their social position.
These may include:
Caste and socioeconomic status.
Race or skin tone.
Disability.
Language.
Age.
Sexual orientation.
Rural or urban location.
Not every characteristic will be relevant to every system. Researchers should identify the characteristics that matter in the particular social and technological context.
7.2 Why is examining gender alone insufficient?
Imagine two women using the same online recruitment platform.
Add to Favorites
Woman A: Rural, low-income and disabled
She may encounter limited internet connectivity, an inaccessible application interface, language barriers or difficulty obtaining technical assistance.
Add to Favorites
Woman B: Urban, affluent and highly educated
She may have reliable internet, stronger digital skills, accessible transport and greater access to technical support.
These are illustrative profiles, not generalizations about all rural or urban women.
Both women may experience gender discrimination. However, their ability to access the system, use it successfully and challenge an incorrect decision may differ.
Research by Sambasivan and colleagues emphasizes that algorithmic fairness must be understood in relation to local socioeconomic conditions and the communities affected by AI.
7.3 How can different inequalities interact?
| Social dimension | Possible barrier in a technological system | Feminist question |
|---|---|---|
| Gender | Historical employment data may reflect unequal opportunities. | Does the system reproduce existing gender disadvantages? |
| Caste | Educational or employment records may reflect caste-based inequalities. | Could the system reproduce unequal opportunities or use proxies associated with social disadvantage? |
| Class | Limited resources may restrict access to devices or connectivity. | Can people with fewer resources use the system effectively? |
| Race or skin tone | Unequal representation may affect model performance. | Does the system perform comparably across relevant groups? |
| Disability | An interface may not support assistive technologies. | Can disabled users access and use the system? |
| Location | Connectivity, language and service availability may vary geographically. | Does the system work in the settings where people actually live? |
These are potential mechanisms to investigate, not proof that every system exhibits each form of bias. NIST specifically recommends considering intersecting demographic groups, data representativeness and the possible use of proxy variables that can indirectly reflect group membership.
7.4 Academic example: The Gender Shades study
The research by Joy Buolamwini and Timnit Gebru provides a concrete example of intersectional algorithmic bias.
Their 2018 study examined three commercial gender-classification systems and found substantial differences in error rates across groups defined by gender and skin tone. The evaluated systems performed worst on darker-skinned female faces.
Consider what researchers might find by examining one category at a time:
Gender alone: They may identify a difference between female and male faces.
Skin tone alone: They may identify a difference between lighter and darker skin tones.
Gender and skin tone together: They may uncover a particularly high error rate for darker-skinned women.
The third analysis reveals a disparity that broader averages can conceal.
This is the practical value of intersectionality: it can reveal how overlapping social categories relate to different outcomes.
7.5 Intersectional fairness in AI research
Academic researchers have also attempted to translate intersectionality into technical methods for evaluating AI systems.
The paper “Differential Fairness: An Intersectional Framework for Fair AI” develops mathematical criteria for evaluating fairness across overlapping social categories. Its authors discuss applications involving census data, criminal recidivism, hospital admissions and loan applications. They also explain that assessing many intersecting groups can be statistically challenging because smaller groups may have limited data.
This limitation is important. Intersectional fairness does not mean collecting every possible personal characteristic indiscriminately. It means carefully identifying relevant disparities, protecting privacy, involving affected communities and interpreting results in context.
7.6 Why is intersectionality especially important in India?
India's social context makes it important to consider how gender interacts with caste, class, disability, language, religion and regional inequality.
Sambasivan and colleagues' 2021 study, “Re-imagining Algorithmic Fairness in India and Beyond,” argues that conventional approaches to algorithmic fairness may overlook local histories and socioeconomic conditions. Based on their interviews and analysis, the authors advocate recontextualizing data and models and empowering marginalized communities to participate in fairer AI ecosystems.
For example, when evaluating a digital recruitment system in India, researchers might investigate whether it is accessible across languages and disability needs, whether its selection criteria reproduce existing educational inequalities, and whether its outcomes differ across relevant social groups.
These questions must be investigated empirically; they should not be treated as proof that a particular system discriminates against any group.
Feminist interpretation: Intersectionality shifts the analysis from asking whether women are disadvantaged to asking which women may be disadvantaged, in what circumstances, through which mechanisms, and with what consequences.
Together, these questions demonstrate why algorithmic fairness cannot be judged solely by technical accuracy. NIST's socio-technical approach and research on intersectional fairness emphasize that responsible evaluation must consider both how a system performs and how its design and deployment affect different people.
The central feminist principle is that technology should not simply reproduce the existing distribution of social power. It should be evaluated according to whether it respects human dignity, expands equal opportunities and makes institutions accountable to the people affected by their decisions.
2. DATA FEMINISM
2.1 Meaning
Data feminism applies feminist and especially intersectional feminist thinking to data science.
Catherine D'Ignazio and Lauren Klein argue that data is not simply an objective reflection of reality. Data is produced through processes involving power, classification, measurement, labour and institutional choices. (Data Feminism)
Central idea
“Who counts” and “what gets counted” are political questions.
For example, unpaid domestic work may be absent or inadequately represented in conventional economic statistics.
But the absence of data does not mean the absence of labour.
2.2 Seven Principles of Data Feminism
D'Ignazio and Klein identify seven major principles. (Data Feminism)
1. Examine Power
Ask:
Who has power over data?
Who:
collects it?
owns it?
analyses it?
controls access?
benefits from it?
2. Challenge Power
Data should not simply describe inequality.
It can also be used to challenge unequal structures.
Example:
Women's organizations can use data on:
gender violence,
wage gaps,
unpaid care work,
political representation
to demand policy change.
3. Elevate Emotion and Embodiment
Traditional data science may privilege numbers over lived experience.
Data feminism argues that people's:
emotions,
bodily experiences,
memories,
lived realities
can also constitute meaningful knowledge.
4. Rethink Binaries and Hierarchies
Feminist analysis questions simplistic classifications such as:
male/female
productive/unproductive
formal/informal
skilled/unskilled
These categories can hide complex realities.
5. Embrace Pluralism
No single dataset or perspective necessarily captures the whole truth.
Data feminism values multiple forms of knowledge, including:
local knowledge,
Indigenous knowledge,
experiential knowledge,
community knowledge.
6. Consider Context
Data cannot always be interpreted meaningfully without understanding the social context in which it was produced.
For example:
“Women participate less in the labour force.”
A feminist researcher asks:
Why?
Possible explanations may include:
unpaid care work,
social norms,
mobility restrictions,
workplace discrimination,
lack of childcare,
safety concerns.
7. Make Labour Visible
Technology may appear automated, but behind many “automated” systems are human workers.
For example:
data labelling,
content moderation,
transcription,
microtasking,
platform customer service.
Data feminism therefore asks:
Who is doing the invisible labour that makes technological systems possible?
These seven principles provide a powerful bridge between feminist theory and technology studies. (Data Feminism)
UNIT 2: THE DIGITAL BODY AND BIOTECHNOLOGY
1. CYBORG FEMINISM
1.1 What is a Cyborg?
Cyborg = Cybernetic + Organism
It refers to a hybrid of:
human/organism + machine/technology
Donna Haraway's famous “A Cyborg Manifesto”, first published in 1985, used the cyborg as a feminist theoretical figure. (Literature Atlas)
1.2 Haraway's Main Argument
Haraway challenges rigid boundaries such as:
Human / machine
Human / animal
Nature / culture
Physical / technological
Her cyborg is therefore not merely a science-fiction robot.
It is a way of thinking about how human identities and bodies are already intertwined with technology.
Examples today
smartphones,
pacemakers,
prosthetic limbs,
cochlear implants,
insulin pumps,
neural interfaces,
wearable health technologies.
Thus:
Technology is increasingly part of how human beings live, communicate and understand their bodies.
1.3 Cyborg Feminism and Gender
Traditional ideas of gender often assume:
Woman = natural body + motherhood + biological identity.
Cyborg feminism questions the idea that technology and biology must remain separate.
It asks:
If technology changes bodies, reproduction and communication, how should we understand gender?
Haraway's argument also challenges the assumption that there is one universal category of “woman.”
This is important for intersectional feminism because women's experiences differ according to:
class,
race,
sexuality,
disability,
nationality,
culture.
1.4 Cyborg Feminism Today: Neural Links and Bio-hacking
Emerging neurotechnology makes Haraway's arguments especially relevant.
Modern neurotechnology can:
measure brain activity,
stimulate the brain,
interact with neural signals,
support treatment of neurological disorders,
potentially interface with computers.
UNESCO notes that neurotechnology raises questions about mental privacy, autonomy, dignity and human identity, particularly when combined with AI. (UNESCO)
Example
A brain-computer interface may potentially allow:
Brain signal → computer system → action
This raises feminist questions:
Who controls the data generated by the brain?
Can employers access neural information?
Could women's emotions or mental states be technologically monitored?
Who gets access to expensive enhancement technologies?
Will technological enhancement create new inequalities?
The OECD similarly identifies concerns around brain-data privacy, human enhancement, commercial manipulation and inequalities in access. (OECD Legal Instruments)
1.5 Bio-hacking
Bio-hacking broadly refers to attempts to modify or optimize biological processes using technological or scientific techniques.
From a gender perspective, the important question is not simply:
“Can we modify the body?”
but:
“Who has the power, resources and right to modify the body—and under what conditions?”
This shifts discussion from technological possibility to:
ethics + power + inequality + autonomy
2. REPRODUCTIVE TECHNOLOGIES
Reproductive technologies include:
IVF,
egg freezing,
sperm donation,
embryo donation,
genetic testing,
gene editing,
surrogacy,
assisted reproductive technologies.
WHO notes that assisted reproduction raises medical, ethical and social questions and has historically generated debates concerning third-party involvement in reproduction, including gametes, embryos and gestation. (World Health Organization)
2.1 CRISPR and Gene Editing
What is CRISPR?
CRISPR-Cas9 is a gene-editing technology that can be used to make targeted changes to DNA.
It has potential applications in:
treating genetic diseases,
developing therapies,
biomedical research.
WHO recognizes the potential of genome editing for diagnosis, treatment and prevention, but stresses that significant ethical, social and governance questions remain. (World Health Organization)
2.2 Somatic vs Germline Editing
Somatic gene editing
Changes are made to cells of an individual and are generally not passed to future generations.
Germline/heritable editing
Changes may affect:
embryo/germline → future child → subsequent generations
This is ethically more controversial because future generations cannot consent to the intervention.
WHO identifies heritable human genome editing as raising particularly significant safety and ethical concerns. (World Health Organization)
2.3 Major Ethical Questions of CRISPR
1. Safety
What if unintended genetic changes occur?
2. Consent
Can future generations consent to genetic modifications?
3. Equality
Will gene-editing technologies be accessible only to wealthy populations?
4. Disability rights
Could attempts to eliminate certain genetic characteristics reinforce the idea that some forms of life are less valuable?
5. “Designer babies”
If technology moves from treating disease to selecting desired characteristics, questions arise concerning:
intelligence,
appearance,
sex,
physical traits.
6. Intergenerational justice
Present generations may make decisions affecting people who do not yet exist.
WHO therefore calls for robust governance and international oversight rather than leaving genome editing entirely to individual markets or laboratories. (World Health Organization)
2.4 Social Egg Freezing
Oocyte cryopreservation means freezing and storing a woman's eggs for possible future use.
The American Society for Reproductive Medicine considers planned oocyte cryopreservation ethically permissible but stresses the importance of informed decision-making because uncertainties remain concerning effectiveness and long-term outcomes. (ASRM)
Why do women freeze eggs?
Possible reasons include:
delaying childbearing,
career planning,
education,
lack of a suitable partner,
medical reasons,
preservation before treatments that may damage fertility.
Feminist Debate Around Egg Freezing
Argument in favour
It may increase:
reproductive autonomy + temporal flexibility
A woman may gain greater control over when she attempts pregnancy.
Critical argument
Individual choice occurs within social structures.
For example:
“Women can simply freeze their eggs and focus on their careers.”
may shift responsibility from institutions to women.
Instead of asking:
Why are workplaces structured around a male breadwinner/continuous-career model?
society may tell women:
“Technology can solve your reproductive problem.”
Therefore feminist analysis asks whether reproductive technologies increase autonomy or sometimes individualize structural inequalities.
2.5 Commercialization/Commodification of the Womb
This refers particularly to debates around commercial surrogacy and the possibility of treating reproductive capacity as a market service.
A feminist analysis asks:
Can pregnancy be treated as labour or a commodity?
Arguments against commercialization
Critics raise concerns about:
exploitation of economically vulnerable women,
bodily risks,
unequal bargaining power,
reproductive labour becoming commodified,
pressure on poor women to become surrogates.
Arguments supporting reproductive choice
Supporters may emphasize:
bodily autonomy,
reproductive freedom,
economic opportunity,
the right to enter reproductive arrangements voluntarily.
Thus, the central tension is:
Autonomy vs exploitation
or:
Choice vs structural inequality
Indian Context
India's Surrogacy (Regulation) Act, 2021 regulates surrogacy and defines commercial surrogacy in terms of payment, reward, fees or monetary incentives beyond permitted medical and insurance expenses. (India Code)
The law establishes a regulatory framework and permits altruistic surrogacy under specified conditions.
Indian bioethics scholarship has highlighted the tensions between:
women's reproductive autonomy,
exploitation,
social stigma,
family structures,
economic inequality,
regulation.
(IMSEAR)
Important conclusion
Reproductive technology is neither automatically liberating nor automatically oppressive. Its impact depends on who controls the technology, who bears the risks, who receives the benefits and what social conditions surround the choice.
UNIT 3: LABOR, PLATFORM ECONOMY, DIGITAL SPACE AND ONLINE RESISTANCE
1. THE GIG ECONOMY
1.1 Meaning
The gig economy refers broadly to labour organized through short-term, task-based or platform-mediated work rather than conventional long-term employment.
Examples include:
Uber/Ola drivers,
food-delivery workers,
online freelancers,
domestic-service platforms,
online tutors,
microtask workers.
The ILO identifies digital labour platforms as an important transformation of contemporary work, creating new opportunities while also producing significant decent-work deficits. (International Labour Organization)
1.2 Gendered Dimensions of Platform Labour
Platform work may provide women:
flexibility,
supplementary income,
access to employment,
opportunities to work from home,
lower entry barriers in some occupations.
But women may simultaneously face:
safety concerns,
unpaid care responsibilities,
digital exclusion,
limited mobility,
lower earnings,
harassment,
weak social protection.
The ILO's research specifically examines women's platform work in India and other countries, including domestic work, tutoring, health services, freelancing, microtasks and e-commerce. (International Labour Organization)
1.3 Gender and Uber/Platform Driving
Recent research on India and Indonesia found very low female participation in platform driving and delivery work. Among women who participate, flexibility is important, but gender differences remain in hours, earnings and exposure to harassment. (National Bureau of Economic Research)
This demonstrates an important point:
Flexibility does not automatically mean equality.
For a woman with childcare responsibilities, flexible work can be beneficial.
But if she can work only during limited hours because of:
household responsibilities,
safety concerns,
restrictions on mobility,
then the same flexibility can produce lower income.
1.4 Amazon Mechanical Turk and Micro-tasking
Amazon Mechanical Turk (MTurk) is an online microtask platform where workers perform small digital tasks.
Examples:
data categorization,
image identification,
transcription,
surveys,
content-related tasks.
Research on MTurk has found gender differences in online labour outcomes. A recent study using data from approximately two million tasks found that women earned less per hour on average, with differences in working patterns—particularly interruptions—playing an important role; the differences were concentrated among workers with children. (MIT Press Direct)
Research specifically examining Indian women workers on MTurk has also connected platform work with household structures, economic necessity and the disciplining effects of platform rules. (SSRN)
1.5 OnlyFans as Gendered Platform Labour
OnlyFans is an interesting example because it demonstrates that platform labour can involve not only conventional services but also digital sexual/intimate labour.
Recent research identifies OnlyFans content creation as a form of digital labour involving:
content production,
self-branding,
marketing,
audience management,
emotional/relational labour,
platform dependence.
Feminist question
Is this:
empowerment through economic independence
or:
commodification of women's bodies and sexuality?
A feminist analysis should avoid assuming either answer automatically.
The relevant questions are:
Who controls the content?
Who receives the income?
What forms of unpaid labour are involved?
What risks do workers face?
How does platform governance operate?
How much autonomy does the worker actually possess?
1.6 Platform Economy: Main Gender Contradiction
Technology promises:
Flexibility → autonomy → income → empowerment
But structural reality may produce:
Flexibility → irregular work → insecurity → unpaid care burden → lower earnings
Therefore:
Key formula
Digital flexibility ≠ automatically gender equality
The ILO similarly emphasizes that women can gain access to platform work while continuing to face uncertainty and barriers to skills development and decent working conditions. (International Labour Organization)
2. AUTOMATION AND DISPLACEMENT
2.1 What is Automation?
Automation means using technology to perform tasks that were previously performed by humans.
AI-based automation can affect:
data entry,
customer service,
administrative work,
accounting,
document processing,
translation,
some professional tasks.
But an important distinction is:
Task automation ≠ complete occupation disappearance
The ILO emphasizes that many jobs are more likely to be transformed or augmented rather than completely eliminated. (International Labour Organization)
2.2 Why is Automation Gendered?
Women are disproportionately represented in many clerical, administrative and business-support occupations.
The ILO's 2026 research finds that female-dominated occupations are almost twice as likely to be exposed to GenAI as male-dominated occupations:
Female-dominated occupations: 29% exposure
Male-dominated occupations: 16% exposure
The ILO links this to women's concentration in clerical, administrative and routine occupations. (International Labour Organization)
Examples include:
secretaries,
receptionists,
payroll clerks,
accounting assistants,
administrative support workers.
2.3 “Pink-Collar” Work
Pink-collar work traditionally refers to occupations culturally associated with women, particularly sectors such as:
clerical work,
care work,
secretarial work,
service work,
some education and health occupations.
The technological risk is not uniform.
Higher exposure
Routine and highly codifiable tasks:
data entry → document processing → routine administrative work
Lower/less direct substitution
Jobs involving:
complex interpersonal relationships,
empathy,
physical care,
judgment,
leadership,
social interaction.
The ILO's evidence stresses that high exposure does not automatically mean complete job loss; it can instead mean that tasks within an occupation are transformed. (International Labour Organization)
2.4 Why the Gender Question Matters
Consider two workers:
Worker A
Male-dominated technical occupation
→ gains access to new AI tools
→ productivity increases
Worker B
Female-dominated clerical occupation
→ routine tasks automated
→ fewer tasks available
→ requires reskilling
Therefore:
The same technological revolution can create opportunities for some workers while increasing insecurity for others.
This is why feminist political economy asks who receives:
the benefits of technological change
and who bears:
the costs of technological change.
3. CYBER-VIOLENCE
3.1 Meaning
Cyber-violence / technology-facilitated gender-based violence refers to violence or abuse against women and girls facilitated through digital technologies.
UN Women identifies technology-facilitated violence as an emerging form of violence against women and stresses the need for better definitions, measurement and policy responses. (UN Women Data Hub)
3.2 Major Forms
1. Cyberstalking
Repeated digital monitoring or unwanted contact.
2. Online harassment
Abusive or threatening communication.
3. Doxxing
Publishing someone's private information—such as an address or other identifying details—without consent.
4. Non-consensual intimate-image sharing
Sharing intimate images without the person's consent.
5. Deepfakes
AI-generated or manipulated:
images,
videos,
audio,
that falsely represent a person.
6. Gendered disinformation
False information specifically designed to attack women because of their gender.
7. Online sexual harassment
Sexualized messages, threats or unwanted sexual content.
The United Nations and UN Women identify these forms as important dimensions of digital violence against women and girls. (United Nations)
3.3 Why Deepfakes Are Gendered
Deepfake technology can be used to place a person's face into fabricated sexual or humiliating material.
This creates several harms:
Privacy harm + reputational harm + psychological harm + social exclusion
Women in public life—including:
politicians,
journalists,
activists,
may be particularly vulnerable to gendered deepfake attacks.
UN Women has highlighted deepfake sexual imagery as a form of technology-facilitated violence that can silence women and discourage their participation in public and political life. (UN Women Knowledge Hub)
3.4 Doxxing and Gender
Doxxing is particularly dangerous because digital information can connect online activity to the physical world.
For example:
Online disagreement
↓
Personal information exposed
↓
Home/workplace identified
↓
Threats or physical intimidation
Thus:
The online/offline distinction is increasingly inadequate for understanding gender violence.
UN sources emphasize that digital abuse can spill over into physical spaces and affect women's safety and participation. (Home | UN Women Transparency Portal)
3.5 Feminist Understanding of Cyber-Violence
Cyber-violence is not simply a “technology problem.”
It is connected to:
patriarchy,
misogyny,
unequal power,
sexual objectification,
political exclusion,
anonymity,
platform design,
weak accountability.
Therefore:
Technology can reproduce offline gender inequalities in digital spaces.
At the same time, technology can also be used to resist those inequalities.
4. DIGITAL ACTIVISM
4.1 Meaning
Digital activism refers to the use of:
social media,
websites,
hashtags,
messaging platforms,
online petitions,
digital campaigns,
video,
digital storytelling
for political and social mobilization.
4.2 Feminist Digital Activism
Examples include:
#MeToo
#NiUnaMenos
#BringBackOurGirls
#EverydaySexism
UN Women identifies the rise of digital activism as an important development in contemporary women's movements. (UN Women Interactive)
UN Women has also highlighted how digital activism and movements such as #MeToo can amplify collective action and increase accountability. (UN Women Data Hub)
4.3 How Technology Helps Feminist Movements
1. Visibility
Previously marginalized voices can reach large audiences.
2. Networking
Activists from different countries can communicate.
3. Documentation
Digital platforms can document:
violence,
discrimination,
human-rights violations.
4. Rapid Mobilization
A hashtag can connect thousands or millions of people around an issue.
5. Agenda Setting
Online discussions can influence:
media,
governments,
institutions,
public opinion.
UN Women notes that digital tools have helped sustain feminist activism and reach groups that might otherwise remain excluded. (UN Women Data Hub)
4.4 Encrypted Communication and Feminist Activism
Encrypted messaging can potentially provide activists with greater privacy when communicating about:
political organizing,
gender violence,
sensitive personal experiences,
protest coordination.
But digital security is not absolute.
Activists can still face:
surveillance,
account hacking,
doxxing,
harassment,
misinformation,
platform censorship.
Therefore:
Digital technology creates both new opportunities for resistance and new mechanisms of control.
This dual character is central to feminist technology studies.
UNIT 3 — OVERALL CONCEPTUAL FRAMEWORK
The four themes can be connected through one major argument:
TECHNOLOGY IS NOT NEUTRAL
Algorithmic AI
↓
May reproduce social bias
Digital body
↓
Technology changes how we understand identity and embodiment
Biotechnology
↓
Creates new possibilities—and new ethical inequalities
Platform economy
↓
Creates flexibility but can reproduce gendered labour inequalities
Cyber-space
↓
Can become a site of both violence and resistance
Digital activism
↓
Technology becomes a tool of feminist mobilization.
A VERY IMPORTANT CONCEPTUAL CHART
| Technology | Potential Benefit | Gender Risk | Feminist Question |
|---|---|---|---|
| AI algorithms | Efficiency | Coded bias | Who designed the algorithm? |
| Facial recognition | Identification/security | Unequal error rates | Whose face is recognized accurately? |
| Hiring AI | Faster recruitment | Employment discrimination | What definition of “merit” is being automated? |
| Data science | Better evidence | Invisible groups | Who gets counted? |
| Neurotechnology | Medical treatment | Mental-privacy risks | Who owns neural data? |
| CRISPR | Disease treatment | Genetic inequality | Who decides what should be edited? |
| Egg freezing | Reproductive planning | Individualization of structural problems | Is technology expanding or commercializing choice? |
| Surrogacy | Parenthood/reproductive opportunity | Exploitation/commodification | Is reproductive labour being commodified? |
| Gig platforms | Flexible employment | Insecurity and care burden | Does flexibility produce equality? |
| AI automation | Productivity | Greater exposure in female-dominated clerical work | Who loses and who gains? |
| Deepfakes | Creative applications | Gendered sexual abuse | How can digital dignity be protected? |
| Social media | Activism and visibility | Harassment/disinformation | Can women participate safely? |
SUGGESTED READINGS IN THE SYLLABUS
I also checked the two suggested readings visible in the image.
1. Zubeeda Banu
The syllabus gives the title approximately as:
“The Impact of New Economic Policy on Women’s Livelihoods – A Case Study From Indian Scenario.”
This publication is traceable as:
Banu, Zubeeda. “The Impact of New Economic Policy on Women’s Livelihoods – A Case Study from Indian Scenario.” Journal of Human Ecology, vol. 10, no. 4, 1999, pp. 265–272.
Bibliographic records confirm the article and its publication details. (Kre Publishers)
Its central relevance to this syllabus is the relationship between:
economic policy → globalization → women's livelihoods → rural/tribal women → environmental and socioeconomic vulnerability
This reading can therefore be connected to the later Unit 3 discussion of gendered labour and economic restructuring.
2. Sahoo and Rath
The second suggested reading is authentic and traceable:
Sahoo, Adyasha, and Navaneeta Rath. “Feminization of Water Collection and Marginalization of Women: The Realities and Reasons.” Academia Environmental Sciences and Sustainability, vol. 2, no. 1, 2025. DOI: 10.20935/AcadEnvSci7504.
The article examines women's disproportionate responsibility for water collection in water-deficit areas of Odisha and links this burden to women's economic participation and quality of life. (Directory of Open Access Journals)
This reading is particularly useful for connecting:
gender + environment + unpaid labour + technology/infrastructure + marginalization
CORE AUTHENTIC SOURCES FOR STUDENTS
For teaching this paper, I would prioritize these sources over random internet material:
NIST — AI Risk Management Framework and AI Bias research: AI bias, fairness, algorithmic decision-making. (NIST)
NIST — Face Recognition Vendor Test: demographic differences in facial recognition. (NIST)
UNESCO — Recommendation on Ethics of AI: gender equality, non-discrimination, transparency and human rights. (UNESCO)
D'Ignazio & Klein — Data Feminism: feminist and intersectional approach to data science. (Data Feminism)
Haraway — “A Cyborg Manifesto”: foundational text for cyborg feminism. (Íñigo Medina)
UNESCO — Ethics of Neurotechnology: neural interfaces, mental privacy, autonomy and dignity. (UNESCO)
WHO — Human Genome Editing: CRISPR, germline editing, ethics and governance. (World Health Organization)
ASRM — Planned Oocyte Cryopreservation: ethical issues surrounding egg freezing. (ASRM)
India Code — Surrogacy (Regulation) Act, 2021: Indian legal framework for surrogacy. (India Code)
ILO — Digital Labour Platforms: gig economy and women's platform work. (International Labour Organization)
ILO — GenAI and Occupational Segregation: gendered impact of automation. (International Labour Organization)
UN Women — Technology-Facilitated Violence: cyberviolence, deepfakes, doxxing and online harassment. (UN Women Data Hub)
UN Women — Digital Activism: feminist movements and digital mobilization. (UN Women Data Hub)
OECD — Responsible Innovation in Neurotechnology: governance, privacy and human enhancement. (OECD Legal Instruments)
conclusion
Gender and Emerging Technology is not simply the study of how women use technology; it is the study of how technology itself is shaped by gendered structures of power—and how technology can either reproduce, challenge or transform those structures.
This framework will help you to move from descriptive answers (“AI is useful/dangerous”) to analytical answers (“How do power, gender, class, labour and inequality shape the design and consequences of AI?”).
Comments
Post a Comment