रवि एक वरिष्ठ पुलिस अधिकारी है, जिसे दंगा-नियंत्रण एवं साइबर-पुलिसिंग का विपुल अनुभव है। पिछले एक वर्ष से वह एक ऐसे ज़िले का पुलिस अधीक्षक (एस॰ पी॰) है, जिसका लगातार दंगों का इतिहास है।
पिछले वर्ष रवि ने अनुमानात्मक पुलिसिंग के लिए एक ए॰ आइ॰ (AI) सॉफ्टवेयर अपने ज़िले में लगवाया। यह प्रणाली लगभग छः महीनों से काम कर रही है। इस नई प्रणाली में भीड़ में मौजूद लोगों के बायोमेट्रिक आँकड़ों को संकलित करने और उन्हें शीघ्रता से एक डेटा भंडार से संबद्ध करने के लिए उन्नत ऐल्गोरिदम का उपयोग किया जाता है। इसने कई तरह के अपराधों में शामिल लोगों को चिह्नित करने में पुलिस की मदद की है।
इस प्रणाली ने प्रवासी तथा निम्न-आय वर्ग के एक आवासीय क्षेत्र को सामूहिक हिंसा और नशीले पदार्थों की तस्करी के केंद्र के रूप में चिह्नित किया है। इस AI प्रणाली की सहायता से स्थानीय पुलिस ने अपनी गश्त, निवारक नज़रबंदी (प्रिवेन्टिव डिटेन्शन) और पुलिस नाके की व्यवस्था केन्द्रित की है। इसके फलस्वरूप सार्वजनिक व्यवस्था और कानून प्रवर्तन में सुधार स्पष्ट रूप से दिखाई देता है।
पिछले सप्ताह कुछ लोक-नेता, नागरिक अधिकारों के वकील, और मानवाधिकार कार्यकर्ता रवि से उसके कार्यालय में मिले। उन्होंने रवि को एक ज्ञापन दिया कि यह नई प्रणाली त्रुटिपूर्ण है, क्योंकि यह गलत ऐतिहासिक तथ्यों पर आधारित है, और सामाजिक पूर्वाग्रहों तथा भेदभावपूर्ण पुलिसिंग से प्रेरित है। ज्ञापन ने यह भी आरोप लगाया कि बढ़ी हुई निगरानी के कारण वहाँ के रहने वालों में एक तनाव व्याप्त है। यह भाव और बढ़ गया है, क्योंकि उस क्षेत्र में रहने वालों को नहीं पता कि किस प्रकार का डेटा उनके नाम से उसमें दर्ज है।
(a) डेटा-आधारित पुलिसिंग में, AI के उपयोग में शामिल नैतिक मुद्दे और पूर्वाग्रह क्या-क्या हैं? (b) रवि के स्थान पर खुद को रखकर विचार कीजिए कि आपके समक्ष कौन-से विकल्प उपलब्ध हैं। यह बताइए कि चयनित विकल्प किस तरह नैतिकता के अधिक निकट है।
Read this question in English
Ravi is a senior police officer with vast experience in riot control and cyber-policing. Since one year, he has been the Superintendent of Police (SP) of a district with a history of frequent rioting.
Last year, Ravi had sought installation of an AI enabled software for predictive policing. This system has been operational for approximately six months. This new system employs advanced algorithms for capturing the biometric data of persons in a crowd and swiftly relating it to a data library. This has enabled the police to identify the persons involved in various crimes.
The system has identified an immigrant and low-income neighbourhood as a centre for gang violence and drug trafficking. Aided by this AI analysis, the local police has focused its patrolling, preventive detentions and establishing checkposts. Consequently, public order and law enforcement has visibly improved.
Last week, some community leaders, civil rights lawyers and human rights activists visited Ravi’s office. They submitted a memorandum that the new system is faulty as it is based on incorrect historical data caused by social biases and discriminatory policing. The memorandum also alleges that the increased surveillance has created a climate of tension amongst residents. This feeling is aggravated by the fact that the residents are not aware of the data noted against their names.
(a) What are the ethical issues including biases involved in the use of AI in data-driven policing? (b) Place yourself in Ravi’s role and discuss the alternatives available. Justify the action that optimises compliance with ethics.
🧭 How to think about this question · directive: “what / discuss / justify”
Decode the demand
- Three directives: part (a) asks 'what' are the ethical issues and biases in AI-driven policing, a listing with reasons; part (b) asks you to 'discuss' Ravi's alternatives and then 'justify' the one that best satisfies ethics.
- The dilemma in one line: public order has visibly improved, but the tool may be punishing a poor, immigrant locality for where the police looked in the past, not for what its residents did.
- The trap: either defending the system because it works, or scrapping it because activists object. The examiner wants a middle course built on verification, audit, transparency and community trust.
- Your seat: you are the SP who asked for the system; your ethics include owning its flaws, not defending your decision.
Think in a structure
- Open with what predictive policing is and the paradox of order gained and fairness lost; a stakeholder hub shows you see every interest.
- Part (a): six crisp points, each led by an ethical keyword (equality, privacy, due process, dignity, accountability) and anchored in a rule or thinker: Articles 14 and 15, Puttaswamy, Kant, Nolan, with NITI, NIST and the EU as evidence.
- Part (b): an options table with merits and demerits, then first-person measures sequenced by time, and a 'why' point that names the ethical tests the chosen course passes.
- End positively: long-term safeguards any SP can propose, then a constitutional-morality anchor from Ambedkar.
Facts checked at source (7)· open
- MeitY lists bias, discrimination, exclusion and lack of transparency among the risks the AI governance guidelines seek to address.PIB — MeitY on India AI Governance Guidelines (19 December 2025) ↗“Some of these include bias, discrimination, unfair outcomes, exclusion, and lack of transparency.”
- NITI Aayog, Responsible AI (2021), Principle of Inclusivity and Non-discrimination: AI should not deepen historic and social divisions based on religion, race, caste, sex, descent, place of birth or residence.Responsible AI #AIForAll: Approach Document for India, Part 1 — NITI Aayog, February 2021 ↗“Principle of Inclusivity and Non-discrimination: AI systems should not deny opportunity to a qualified person on the basis of their identity. It should not deepen the harmful historic and social divisions based on religion, race, caste, sex, descent, place of birth or residence in matters of education, employment, access to public spaces, etc.”
- EU AI Act: real-time and remote biometric identification, such as facial recognition in public spaces, is among banned AI applications, with limited exceptions for law enforcement.EU AI Act: first regulation on artificial intelligence — European Parliament ↗“Banned AI applications in the EU include: Cognitive behavioural manipulation of people or specific vulnerable groups: for example voice-activated toys that encourage dangerous behaviour in children Social scoring AI: classifying people based on behaviour, socio-economic status or personal characteristics Biometric identification and categorisation of people Real-time and remote biometric identification systems, such as facial recognition in public spaces Some exceptions may be allowed for law enforcement purposes.”
- Kant's Humanity Formula: never treat humanity, in oneself or in others, as a means only but always as an end in itself.Kant's Moral Philosophy — Stanford Encyclopedia of Philosophy ↗“This formulation states that we should never act in such a way that we treat humanity, whether in ourselves or in others, as a means only but always as an end in itself.”
- NIST (US), December 2019: the majority of face recognition algorithms exhibit demographic differentials, i.e. their ability to match two images of the same person varies from one demographic group to another (FRVT Part 3, NISTIR 8280).NIST Study Evaluates Effects of Race, Age, Sex on Face Recognition Software, National Institute of Standards and Technology, 19 December 2019 ↗“the majority of face recognition algorithms exhibit demographic differentials. A differential means that an algorithm’s ability to match two images of the same person varies from one demographic group to another.”
- DPDP Act, 2023, Section 17(1)(c): Chapter III (rights and duties of data principals) does not apply where personal data is processed for the prevention, detection, investigation or prosecution of any offence.The Digital Personal Data Protection Act, 2023 (No. 22 of 2023), Ministry of Electronics and Information Technology ↗“17. (1) The provisions of Chapter II, except sub-sections (1) and (5) of section 8, and those of Chapter III and section 16 shall not apply where— … (c) personal data is processed in the interest of prevention, detection, investigation or prosecution of any offence or contravention of any law for the time being in force in India;”
- Rawls: the veil of ignorance keeps arbitrary facts about citizens out of the choice of principles of justice; a citizen's race, class or gender is no reason for social institutions to favour or disfavour her.John Rawls, Stanford Encyclopedia of Philosophy ↗“The most striking feature of the original position is the veil of ignorance, which prevents arbitrary facts about citizens from influencing the agreement among their representatives. As we have seen, Rawls holds that the fact that a citizen is of a certain race, class, and gender is no reason for social institutions to favor or disfavor her.”
Question: UPSC's CS (Main) 2026, GS Paper IV — paper ↗. Guidance: Minimalist IAS, checked 8 Nov 2026 (how we verify). ·
Model answer· 574words · Minimalist IAS
Predictive policing forecasts crime from past records. In Ravi's district the system has visibly improved order, yet it has marked an immigrant, low-income locality as suspect; trained on arrest data, it learns where the police looked, not where crime is. That tension frames both parts.
Stakeholders: residents (dignity, privacy, equality); victims of gang crime (safety); Ravi's force (legitimacy, discipline); community leaders and rights groups (accountability); the vendor (liability)
(a) Ethical issues, including biases, in AI-driven policing
- Feedback-loop bias: skewed historical data sends more patrols to one area; their fresh records confirm the pattern, offending Articles 14 and 15.
- NITI Aayog's Responsible AI (2021): AI should not deepen historic divisions based on religion, race, caste, descent, place of birth or residence.
- Technical bias: matching accuracy differs across groups, so false matches and wrongful detentions fall unevenly on minorities.
- NIST (2019): most face-recognition algorithms show demographic differentials in matching two images of the same person.
- Privacy and proportionality: mass crowd biometrics without a statute fail Puttaswamy's tests of legality, legitimate aim and proportionality.
- EU AI Act bans real-time remote biometric identification in public spaces, with narrow law-enforcement exceptions.
- Opacity and due process: residents cannot see or contest their records; natural justice is denied and preventive detention rests on an unexplained score.
- Dignity: treating a whole locality as a risk profile treats people as means, not ends, which Kant's humanity formula forbids.
- Automation bias and accountability: officers defer to the screen, yet accountability for liberty lies with the officer, not the algorithm (Nolan principles).
- MeitY's India AI Governance Guidelines (2025) list bias, discrimination, exclusion and lack of transparency as risks to be addressed.
(b) Alternatives before me as SP
| Option | Merits | Demerits |
|---|---|---|
| Continue as is | Visible order; force morale | Entrenches bias; erodes trust; legally fragile |
| Scrap the system | Harm stops; trust restored | Riot risk returns; a useful tool lost |
| Pause high-risk uses, audit and reform | Order with fairness; lawful | Short-term slowdown; vendor resistance |
(b) The action I would take and why it optimises ethics
- Human-in-the-loop at once: no detention, search or checkpost on AI output alone; an officer verifies every match and records reasons.
- Independent audit: data and algorithm examined by outside experts, retraining on cleaned data, SOPs published for scrutiny.
- Transparency and correction: a grievance officer lets residents know and correct their records, since the DPDP Act's Section 17 exempts crime-related processing from such rights.
- Community policing: joint beat meetings with the leaders who met me, civic services, so order rests on cooperation, not fear.
- Kerala's Janamaithri Suraksha Project is a working model of beat-level police–community partnership.
- Upward transparency: I place the memorandum and my orders before the Collector and DGP, inviting oversight rather than avoiding it.
- Why this course: it retains the legitimate aim, riot prevention, with equality, privacy and due process; Puttaswamy's proportionality and Rawls's justice both demand it.
- Rawls: a citizen's race or class is no reason for institutions to disfavour her; nor is her address.
Long-term safeguards
- A statute on police biometrics: purpose limits, retention periods and judicial oversight meeting Puttaswamy's tests.
- Published bias audits: periodic, independent, with results open to the public and courts.
- Training: officers taught to question machine output and to treat predictions as leads, never as evidence.
Technology should sharpen policing, not decide it. Ambedkar held that constitutional morality must be cultivated, not assumed; a force that lets residents see, question and correct what a machine says of them cultivates it, and earns a public order that lasts.
Prepared by Minimalist IAS to the answer standard of Shivam Singh (mentor, 7 years) · facts checked against official sources (how we verify) · updated 8 Nov 2026. UPSC publishes no model answers: compare your structure and coverage with this, then write your own.
