aiMOOC

ABI - Digital responsibility and ethics

QR-Code

ABI - Digital responsibility and ethics

Mini-description: Discuss case-based questions on privacy, bias, accountability, transparency and digital responsibility from several serious perspectives. You work mainly in English, analyse sources and media, and build your own evidence-based position. The course is designed for Englisch als fortgeführte Fremdsprache, Kursstufe Baden-Württemberg, as a shared Basisfach/Leistungsfach learning path.

Media kick-off: What can a digital system know about you without asking you a direct question? Before watching, predict three kinds of data that could be inferred. While watching, note one claim, one example and one question. After watching, distinguish between a privacy problem, a design choice and a social consequence.

Einleitung

Digital responsibility is not one rule. It is a way of asking who benefits, who may be harmed, who can challenge a decision, what information is collected, and who must answer when a system fails. In this aiMOOC you use five lenses:

Lens	Core question	Typical evidence
Privacy	What personal data is collected, inferred, stored or shared?	data flows, consent, purpose, retention
Bias and fairness	Who may be treated differently, and why?	datasets, error rates, proxies, outcomes
Accountability	Who is responsible for decisions and remedies?	roles, logs, review, appeal
Transparency	What should users and affected people be able to understand?	notices, explanations, limitations, traceability
Digital responsibility	What should people and organisations do before, during and after deployment?	safeguards, monitoring, source checking, human oversight

This is a case-based course. The cases are fictional and use invented names and data. You do not need to disclose personal information, create public recordings, upload faces or voices, or use an external AI tool.

Safe and responsible working rules
Use fictional data in all case work. Recordings are voluntary and stay private unless you explicitly choose otherwise. Do not upload classmates' images, voices, screenshots or personal data. No external AI service is required. If you use AI independently, document the tool, verify factual claims against reliable sources, check quotations and references, and write the final analysis yourself. AI output is not a substitute essay.

I can ...

I can ...	BF focus	LF extension
explain privacy, bias, accountability and transparency in clear English	use one case and one source	connect several cases and perspectives
analyse how a text, chart or video frames digital responsibility	identify claim, evidence and language	compare framing, assumptions and limitations
discuss ethical trade-offs without reducing them to yes/no answers	structured pro/con reasoning	weigh competing principles and consequences
speak and write at B2 level, partly towards C1	sentence frames and guided structure	greater lexical range, abstraction and independent synthesis
use Abitur operators accurately	summarize, analyze, discuss	assess, evaluate and connect materials independently

Level orientation: Baden-Württemberg expects learners in the course stage to reach B2, in parts C1. BF and LF differ especially in complexity, abstraction, breadth, depth and differentiation of materials and tasks.

Abi-Check: current official context

Checked on 3 October 2026. For the Baden-Württemberg Abitur 2027 and 2028 in English, the official Facherlasse continue to base the examination on the Bildungsplan 2016. They name On the Move: Migration and Cross-Cultural Encounters as the Schwerpunktthema and the USA/UK fields Politics, culture, society – between tradition and change. This aiMOOC does not present “digital responsibility” as an additional prescribed Schwerpunktthema. It uses the topic to train the required functional communicative competences, text and media competence, current-issues discussion and Abitur-style operators.

For LF, the 2028 Facherlass specifies listening plus writing, including aspect-based summary, analysis and a choice of personal response or creative target text; it also points to the IQB operator list valid from examination year 2027. For BF, the course and assessment remain differentiated in scope and depth, with all three Anforderungsbereiche represented over the qualification phase.

Always check your own examination year and school information before final exam preparation.

Official starting points:

Vorwissen aktivieren

Look at the image and answer in English: Where does responsibility begin — with the user, the developer, the organisation, the state, or all of them?

Quick check	Your first answer	Revisit after the course
A useful service asks for more data than it needs. What is the issue?
A model is accurate overall but makes more errors for one group. What should be checked?
A person cannot discover why an automated decision affected them. What is missing?
A company says “the algorithm decided”. Is that a complete accountability answer?

Core input: five lenses, five cases

Case 1: Privacy — the study app

Case A — StudyPulse: A fictional school offers an optional revision app. It records quiz results, time-on-task and device type. A new update also collects precise location to “improve study recommendations”, although the recommendation engine does not need location. Data are kept for three years. The privacy notice is long and difficult to understand.

BF task: Outline which data practices need closer scrutiny. Use the concepts purpose, data minimisation and retention.

LF extension: Assess whether consent alone would solve the ethical concerns. Distinguish legal compliance, meaningful choice and good design.

A useful EU perspective: under the GDPR, personal data processing is guided by principles including lawfulness, fairness and transparency, purpose limitation, data minimisation, storage limitation, accuracy, security and accountability. Use the official overview rather than memorising a slogan: European Commission: Data protection explained.

Viewing task: Note two points the European Commission emphasises. Then ask: Which important ethical questions are not answered by a short legal explainer?

Case 2: Bias — the internship filter

Case B — FutureHire: A fictional company uses a model to rank internship applications. The training data come from previous successful applicants. Historically, most interns came from a small set of schools and neighbourhoods. The model does not receive ethnicity, but postcode and school are included. Internal testing shows similar overall accuracy but clearly different false-negative rates across groups.

Listening/Viewing: While watching the Crash Course video, make a five-box note map for possible sources of algorithmic bias. Do not copy full sentences; paraphrase.

BF task: Analyze how historical data and proxy variables could affect FutureHire.

LF extension: Evaluate two possible interventions. For each, state what improves, what may remain unresolved and what new evidence would be needed.

Perspective check: Joy Buolamwini foregrounds lived consequences and accountability. Compare this with a technical fairness perspective: What counts as evidence in each? Where do they overlap?

Case 3: Transparency — the scholarship score

Case C — GrantScore: A fictional foundation uses an AI-assisted score to shortlist scholarship applicants. Rejected students receive only: “Your profile did not meet the threshold.” The vendor says the model is proprietary. The foundation can see a confidence score but not a clear reason for each decision. A student asks for a human review.

Datei:Ethics of AI in Digital Medicine – Explanation why transparent explainable AI is important.webm

Key distinction: Transparency is broader than explaining one prediction. It can include disclosure that AI is used, information about purpose and limits, meaningful explanation for affected people, documentation for auditors and traceability for those responsible.

BF task: Build a three-level explanation: what a user should know, what a decision-maker should know and what an auditor should know.

LF extension: Discuss whether complete technical openness is always necessary for meaningful transparency. Consider privacy, security, intellectual property and the right to challenge a consequential decision.

As of 2026, the EU AI Act contains transparency duties for certain AI systems. The European Commission issued Article 50 guidelines in July 2026, with relevant transparency obligations applying from 2 August 2026. Use the official guidance for current details: European Commission: AI Act transparency guidelines.

Case 4: Accountability — the health chatbot

Case D — CareGuide: A fictional clinic offers a chatbot that gives general preparation advice before appointments. It mistakenly tells a patient to stop a prescribed medicine. The screen includes a small disclaimer. The clinic purchased the tool from a vendor, staff were told to monitor complaints, but no one was assigned to review high-risk messages.

Accountability map:

Actor	Possible responsibility	Evidence to inspect
provider	design, testing, documentation, known limitations	model documentation, tests, incident reports
deploying organisation	purpose, safeguards, human oversight, monitoring	procedures, staff roles, logs
professional user	appropriate use and escalation	training, interface, workflow
affected person	should not carry the burden of system governance	notice, complaint and remedy routes

BF task: Identify where the chain of responsibility broke.

LF extension: Assess how responsibility should be distributed when several actors contributed to the failure. Avoid the empty sentence “everyone is responsible”; specify duties, evidence and remedy.

A risk-management perspective: the US NIST AI Risk Management Framework is voluntary and focuses on managing risks and trustworthiness characteristics such as validity, safety, security, accountability, transparency, explainability, privacy and fairness. It is not an EU law. NIST AI RMF.

Case 5: Digital responsibility — synthetic media

Case E — CampusVoice: A fictional student media team creates an AI-generated audio clip imitating the principal for a satire project. The clip is clearly labelled in the classroom version. A student later reposts a cropped version without the label. Some listeners believe it is authentic.

BF task: Discuss what the creators, the reposter and the platform should do next.

LF extension: Evaluate which preventive measures are proportionate: visible labels, provenance metadata, editorial review, media literacy, access restrictions or other safeguards. Explain trade-offs.

The current EU AI Act includes transparency rules for certain AI-generated or manipulated content. Use the consolidated legal text and current Commission guidance rather than relying on a social-media summary: EUR-Lex: consolidated AI Act.

Serious perspectives: compare, do not collapse

Perspective	Main emphasis	Useful question	Limitation to remember
GDPR / EU data protection	rights and lawful, fair, transparent personal-data processing	Is the data use necessary, clear and accountable?	data protection does not answer every ethical question
EU AI Act	risk-based duties for AI actors and specific transparency requirements	What obligations follow from the system and context?	legal classification is not the same as a complete moral evaluation
UNESCO AI ethics	human rights, dignity, fairness, oversight, sustainability, accountability	Who may be harmed or excluded?	principles still need concrete implementation
OECD AI Principles	trustworthy, human-centred AI, transparency, robustness and accountability	Can affected people understand and challenge outcomes?	broad principles require context-specific evidence
NIST AI RMF	practical risk management across the AI lifecycle	How can risks be mapped, measured, managed and governed?	voluntary framework; not a substitute for applicable law

Useful source set:

Listening / Viewing

Choose one core video and one contrasting video. Use subtitles only after a first listening if possible.

Media	First listening	Second listening	BF / LF
Crash Course: algorithmic bias	identify the main problem	note examples and causal links	BF: 3 points; LF: map 5 bias mechanisms
Joy Buolamwini: coded gaze	identify the speaker's concern	track example, evidence, appeal	BF: summary; LF: rhetorical analysis
IBM: explainable AI	define XAI in context	distinguish explanation from transparency	BF: concept map; LF: limitations
TED: surveillance and privacy	identify thesis	separate factual examples from evaluation	BF: 4 notes; LF: source/framing critique

Listening self-check: Did you record keywords instead of sentences? Did you separate what the speaker says from what you think? Can you identify an example that supports the main claim?

Reading

Micro-reading set — use the linked official pages, not copied full texts.

Source	Reading purpose	BF	LF
European Commission GDPR principles	identify privacy principles	select 3 relevant principles for Case A	connect principles to rights and remedies
UNESCO Recommendation	identify ethical principles	choose 2 principles for one case	compare tensions among 4 principles
OECD AI Principles	track transparency and accountability	paraphrase one principle	compare OECD and UNESCO framing
NIST AI RMF	understand risk management	explain one trustworthiness characteristic	connect lifecycle governance to a case
EU AI Act guidance	identify current transparency duties	locate one relevant rule	distinguish legal duty from broader ethical argument

Reading strategy: Preview headings. Ask one question before reading. Mark claims, definitions and limits. Use a monolingual English dictionary first. Use German only if context, word family and English explanation do not solve the problem.

Kontextwortschatz

Word	English explanation	Word family / collocation	Example	German help only if needed
privacy	control and protection concerning personal information and private life	private, privacy-preserving	Privacy is affected when unnecessary location data are collected.	Privatsphäre / Datenschutzkontext
bias	a systematic tendency that can skew judgement or outcomes	biased, algorithmic bias	A proxy variable may reproduce historical bias.	Verzerrung / Voreingenommenheit
fairness	treatment or outcomes judged to be just in a given context	fair, unfair, fairness metric	Fairness can require more than equal overall accuracy.	Fairness / Gerechtigkeit
accountability	being answerable for actions, decisions and remedies	accountable, hold accountable	Accountability requires clear roles and evidence.	Rechenschaftspflicht
transparency	availability of relevant information about a system or process	transparent, transparently	Transparency should be meaningful for the audience.	Transparenz
explainability	the degree to which reasons or system behaviour can be explained	explainable, explanation	Explainability can help people challenge a decision.	Erklärbarkeit
traceability	ability to reconstruct relevant steps, data or decisions	trace, traceable	Logs can support traceability after an incident.	Nachvollziehbarkeit
oversight	supervision by people or institutions	human oversight, regulatory oversight	Human oversight must be more than a symbolic button.	Aufsicht / Kontrolle
data minimisation	collecting only data necessary for a specified purpose	minimise, minimal	Precise location fails the minimisation test if it is unnecessary.	Datenminimierung
remedy	a way to correct harm or unfair treatment	remedial action, seek a remedy	An appeal process can provide a remedy.	Abhilfe
trade-off	a situation in which improving one value may weaken another	balance a trade-off	More disclosure may create a security trade-off.	Zielkonflikt
stakeholder	a person or group affected by or able to affect a decision	stakeholder analysis	Students are stakeholders in an education platform.	Beteiligte / Betroffene

Word-building drill: explain → explanation → explainable → explainability; account → accountable → accountability; trace → traceable → traceability; discriminate → discrimination → discriminatory; transparent → transparency.

Analyse-Werkzeuge

CLAIM–EVIDENCE–ASSUMPTION–IMPACT

Use this four-step grid for any source or case:

Step	Prompt
Claim	What is being argued or decided?
Evidence	What data, examples, quotations or observations support it?
Assumption	What must be accepted for the claim to work?
Impact	Who benefits, who bears risk and who can contest the outcome?

BF scaffold: one strong point per row, then connect the rows.

LF extension: compare two sources, identify conflicting assumptions and judge which additional evidence would change your conclusion.

Case matrix

Case	Main lens	Secondary lens	Best evidence	AFB II / III move
StudyPulse	privacy	accountability	data categories, purpose, retention	analyse necessity; assess safeguards
FutureHire	bias	transparency	group-level error patterns, proxies	analyse causes; evaluate interventions
GrantScore	transparency	accountability	notice, explanation, review route	examine information gaps; discuss challenge rights
CareGuide	accountability	safety	workflow, monitoring, incident logs	analyse role failures; assess remedies
CampusVoice	digital responsibility	transparency	labels, provenance, repost context	discuss duties; evaluate proportionate measures

Speaking

60-second position: “A responsible digital system should be judged by what happens when it fails, not only by how well it performs when it works.” Prepare a claim, two reasons, one counterargument and a conclusion.

BF support: Use: “My main point is … / The strongest evidence is … / A possible objection is … / However … / Therefore …”

LF extension: Integrate two institutional perspectives and one case. Qualify your claim with language such as “to a large extent”, “in high-stakes contexts”, “provided that”, “unless”.

Dialogue task: In pairs, one student represents the deploying organisation and one represents an affected person. After two minutes, swap roles. Do not role-play real classmates or use real personal data.

Writing

Abitur-style practice, not a replacement essay:

Task A — analyze: Analyze how one source frames responsibility for an AI-related risk. Focus on structure, use of language and communicative strategies.

Task B — assess: Assess to what extent meaningful transparency is necessary for accountability in a high-impact digital system. Refer to at least two course materials.

Task C — target text: Write a concise article for a school technology newsletter explaining three safeguards that should be in place before an AI-supported decision system is introduced.

BF target: 250–350 words with a clear paragraph plan and two materials.

LF target: 400–550 words, three or more materials, explicit comparison of perspectives and a more independent AFB II/III argument.

Writing rule: Plan and draft independently. If you use an AI tool outside this course, do not paste personal or school-internal data. Record the tool and purpose, verify factual statements and source links, and revise every claim in your own reasoning and language.

Mediation / Sprachmittlung

Optional mediation task: A German school committee has a short German-language notice about a planned AI attendance system. An English-speaking exchange student group asks what matters for privacy and accountability. Write an English email that presents the relevant information for this audience. Do not translate sentence by sentence. Select, explain and adapt culturally relevant information.

BF: 180–220 words with a planning grid.

LF: 250–320 words; add a concise explanation of why one issue may be controversial, without adding unsupported facts.

Hilfen

If you are stuck ...	First move	Next move	German fallback
on vocabulary	infer from context and word family	use an English-English dictionary	check a short German gloss
on a source	identify heading, claim and example	mark evidence and limitation	clarify one difficult sentence
on analysis	use claim–evidence–assumption–impact	connect language/form to purpose	restate the task operator
on speaking	use a 4-part note card	add one counterargument	prepare 3 key phrases
on writing	build paragraph claims first	attach evidence before drafting	check task fulfilment in German only if necessary

Feedback und Fehleranalyse

Peer feedback code: C = claim unclear; E = evidence missing; L = link between evidence and claim weak; P = perspective missing; R = register; V = vocabulary; G = grammar; S = source not verified.

After feedback, do not simply “correct mistakes”. Complete this mini-log:

Error / weakness	Why it happened	Better strategy	New example

Quality check: A strong response distinguishes fact from judgement, states whose perspective is being used, uses evidence proportionately and shows what remains uncertain.

Transfer

Choose a new context: smart city, school analytics, recruitment, health, finance, social media, generative AI or public administration. Apply all five lenses. Do not assume that one principle automatically outweighs the others. Ask what changes when stakes, users, data sensitivity and possibility of appeal change.

Abi-Check

Use the current IQB operator meanings from examination year 2027 onward. In short:

Operator	What your response must do
summarize / outline	give a concise account focused on the task
analyze / examine	describe and explain in detail, using evidence and relevant aspects of language/form
discuss	develop arguments and reasons for and against and reach a well-founded conclusion
assess / evaluate	give a well-founded judgement of value, effectiveness or significance, supported by evidence and examples

Abi self-check: Did you answer the exact operator? Did you use material evidence? Did you analyse rather than retell? Did you connect language/form to function? Did you qualify your judgement? Did you reserve time for language revision?

Folgekurs

Continue with ABI - AI, media and democracy, ABI - Data privacy and surveillance, ABI - Algorithmic bias and fairness or ABI - Argumentative writing and source evaluation. For LF, add a cross-source synthesis task with an unfamiliar non-literary text and a chart. For BF, revisit one case with a shorter source and stronger planning support.

Interaktive Aufgaben

Quiz: Teste Dein Wissen

Which concept asks whether only necessary personal data are collected for a stated purpose? (Data minimisation) (!Transparency) (!Traceability) (!Automation)

Which term best describes a systematic tendency that can skew algorithmic outcomes? (Bias) (!Retention) (!Consent) (!Encryption)

What is the strongest example of accountability? (Clear roles plus review and remedy) (!A long disclaimer) (!A high accuracy score) (!A secret model)

What does meaningful transparency aim to provide? (Relevant understandable information for the audience) (!Every line of source code for everyone) (!A promise that errors cannot happen) (!A guarantee of perfect fairness)

Which action is most useful when a model has different false-negative rates across groups? (Investigate data outcomes and possible causes) (!Ignore group patterns if overall accuracy is high) (!Remove all documentation) (!Collect more data without a purpose)

Which statement about the NIST AI Risk Management Framework is correct? (It is a voluntary risk management framework) (!It is an EU criminal law) (!It replaces the GDPR) (!It bans all automated decisions)

Which practice best supports traceability? (Keeping relevant logs and documentation) (!Deleting all records immediately) (!Using vague responsibility labels) (!Hiding the system purpose)

What is a suitable first step in source evaluation? (Identify the claim and the evidence supporting it) (!Accept the headline as proof) (!Translate every word before reading) (!Use only one perspective)

Which action follows the course AI-use rule? (Verify AI claims and write the final analysis yourself) (!Submit an AI-generated replacement essay) (!Upload classmates' personal data for better advice) (!Trust invented citations if they sound plausible)

Which operator requires a well-founded judgement supported by evidence? (Assess) (!Copy) (!List) (!Name)

Memory

Privacy	Protection and appropriate control of personal information
Bias	Systematic distortion in outcomes
Accountability	Answerability for decisions and remedies
Transparency	Relevant information made understandable to an audience
Explainability	Ability to give reasons or make system behaviour comprehensible
Traceability	Ability to reconstruct relevant steps and decisions

Drag and Drop

Ordne die richtigen Begriffe zu.	Thema
Data minimisation	Collect only what is necessary for the specified purpose
Human oversight	Ensure a person can supervise and intervene appropriately
Bias audit	Examine patterns that may disadvantage groups
Appeal route	Allow an affected person to challenge a consequential outcome
Source verification	Check claims and references against reliable evidence

Kreuzworträtsel

Privacy	Which concept concerns protection and appropriate control of personal information?
Bias	Which term describes systematic distortion in decisions or outcomes?
Audit	What one-word process can independently examine a system or procedure?
Consent	What term means a person's informed agreement?
Fairness	Which concept asks whether treatment or outcomes are just in context?
Traceability	What term means the ability to reconstruct relevant steps or decisions?

LearningApps

Lückentext

Digital responsibility asks who is affected and who is

for decisions and remedies.

Collecting only necessary personal data follows the principle of

.

A systematic distortion in outcomes can be described as

.

Information should be meaningful to its intended

.

A person affected by a consequential decision may need a route to

it.

Relevant records can improve

after an incident.

A strong argument connects a claim to supporting

.

AI-generated information should be checked against reliable

before it is used.

Offene Aufgaben

Leicht

Privacy map: Draw a data-flow diagram for StudyPulse using only fictional data. Mark collection, purpose, storage and deletion.
Vocabulary visual: Create a one-page visual connecting privacy, bias, accountability, transparency and traceability with one English example each.
Video note challenge: Watch one embedded video and produce a 60-word English summary plus one open question.
Source check: Compare the title, authoring institution and purpose of two official source pages. Explain which question each source can and cannot answer.

Standard

Case podcast: Record a voluntary private two-minute audio discussion of one case, or submit a written dialogue instead. No real names or personal data.
Bias intervention: Propose two changes to FutureHire and explain what evidence would show whether each change worked.
Transparency notice: Rewrite the GrantScore rejection notice so that it is clearer and more useful without inventing a reason the system did not provide.
Responsibility matrix: Assign concrete duties to provider, deploying organisation and human user in CareGuide. Add one remedy for an affected person.

Schwer

Perspective synthesis: Compare UNESCO, OECD and NIST on one case. Identify one shared principle and one meaningful difference in emphasis.
Regulation and ethics: Explain why legal compliance and ethical responsibility can overlap without being identical. Use one GDPR or AI Act source and one ethics framework.
Mini hearing: Stage a structured hearing on CampusVoice with three roles. Each role must cite evidence and respond to a counterargument.
Abitur transfer essay: Write an independent response to: “Transparency without accountability is only information.” Use at least three materials and include a qualified conclusion.

Lernkontrolle

Case transfer: A fictional university uses AI to flag “low engagement” and contacts students automatically. Analyse the case through privacy, bias and accountability. Then propose two proportionate safeguards.
Material comparison: Compare how an official legal source and an ethics framework define or operationalise responsibility. Explain why the difference matters in practice.
Evidence challenge: A company claims its system is fair because accuracy is 94 percent. Explain why this figure alone is insufficient and state what additional evidence you would request.
Transparency design: Design a two-layer explanation for an AI-assisted decision: one layer for an affected person and one for an internal reviewer. Justify what belongs in each.
Speaking transfer: Give a three-minute response to “Human oversight is meaningful only if a human has the authority, information and time to intervene.” Include a counterargument.
Writing transfer: Assess to what extent digital responsibility should be shared across developers, deployers and users. Use a fictional case and at least two serious perspectives.

Lernnachweis

For a strong learning record, you should be able to show:

Concept knowledge: accurate use of privacy, bias, fairness, accountability, transparency, explainability, traceability and oversight.
Source literacy: ability to distinguish official legal/regulatory material, voluntary frameworks, ethics recommendations, educational videos and commentary.
Analysis: evidence-based interpretation of claims, language, visual material, assumptions and impacts.
Argumentation: a balanced, well-founded conclusion that acknowledges trade-offs and uncertainty.
Language: clear B2 English, partly towards C1, with appropriate register, linking and topic vocabulary.
BF-LF differentiation: BF demonstrates secure core analysis with structured support; LF demonstrates broader source linkage, greater depth and more independent AFB II/III reasoning.
Academic integrity: independent writing, transparent source use, verified references and documented AI assistance if used outside the course.
Privacy practice: fictional data, voluntary private recordings and no unrequested uploads of personal information.

OERs zum Thema

Further open and official resources:

Verknüpfte Lernbereiche

ABI - Digital responsibility and ethics

aiMOOC-Projekte

Inhalte werden geladen ...

Mediathek wird aus dem Wiki geladen ...

Frag mich zum aktuellen Lerninhalt oder wähle eine kurze Hilfe.

Textmodus bereit

Sprache