DC Elective 2: Ethical Issues in Digital Communication

Digital Communication Certificate · Elective 2

Ethical Issues in Digital Communication

This elective applies a structural AI literacy framework to ethical questions in digital communication. Its anchor case is the governance of AI-based academic integrity tools. From there, you will make principled operative decisions about digital tools in your own teaching. Plan on 3 to 4 hours: about 2 on the activities and the builder, the rest reading.

What this elective asks you to do

Analyze a digital tool with the three structural questions. (Activity 1)
Evaluate the decision path taken in a governance case. (Activity 2)
State and justify an operative decision for the tool you analyzed. (Activity 2, prompts 4 and 5)
Design an assignment teaching structural ethical analysis. (Activity 2 and the assignment builder)

Section 1

The AI Literacy Framework as Ethical Procedure

Ethical questions about digital tools turn on what the tools do. A faculty member who adopts a plagiarism detection system to promote academic integrity still answers for how the system's effects fall across students. Ethics in digital communication asks whose interests a system serves, whom it harms, and how it distributes risk. Intentions do not settle those questions.

The AI literacy framework (Cole, 2026) offers a three-level procedure for this kind of structural ethical analysis. The levels run in sequence, and each is a prerequisite for the next. Conceptual literacy comes before structural analysis, and structural analysis comes before operative decision-making. Skipping to the operative level (asking "should we use this tool?" before asking "what does this tool do and whose interests does it reproduce?") produces decisions that mistake good intentions for adequate ethical analysis.

Conceptual literacy establishes accurate naming and description. Before you judge whether a tool is ethical, you need to know what kind of system it is, what its training data and objective function are, and what it claims to detect or produce. Decisions about AI tools go wrong when the people making them treat probabilistic systems as definitive, pattern-matching as semantic comprehension, or training-data outputs as ground truth.

Scope of this elective

This elective sets aside individual-level digital ethics, such as privacy settings, password hygiene, and citing AI-generated content. Those topics matter, and they belong elsewhere. The questions here are structural: what systems are deployed, by whose authority, to whose benefit and whose detriment, and on what grounds.

Section 2

Structural Analysis: Three Questions

The structural level of the AI literacy framework applies three questions to any digital system under consideration. The questions are designed to show what a system reproduces, whatever it claims to do.

Question 1: What hierarchies does this system reproduce? Every AI system is trained on data that encodes existing social arrangements: linguistic norms, institutional practices, patterns of who gets heard and who does not. A natural language processing system trained mostly on Standard American English academic prose will underperform on texts in other registers or dialects and on texts by writers whose first language is not English. A patch will not fix this. The underperformance follows from the training data, and it reproduces the hierarchies encoded in that data.

Question 2: Who bears the costs when the system fails? Failure costs in AI systems rarely fall evenly. When an AI content moderation system produces a false positive, the person whose content was wrongly flagged pays for it. The developer, the institution, and the majority of users for whom the system performs well bear none of that cost. Asking who bears failure costs is an ethical question, and answering it requires knowing whose position leaves them most exposed to the system's errors.

Question 3: Whose interests does the design serve? AI systems are designed to meet the needs of those who fund and commission them. An academic integrity system is designed for institutions seeking to detect misconduct. The students it processes are not the party it was built to serve. That asymmetry leaves open whether the system is legitimate. Any claim about what the system is "for" has to be weighed against whose interests shaped its design.

Section 3

Operative Decisions: Three Paths

After structural analysis, the operative level calls for a decision. The framework identifies three legitimate outcomes, each with its own conditions.

Principled Refusal

The structural analysis reveals harms that cannot be mitigated within the deployment context. A decision not to deploy, grounded in analysis, is itself an ethical act. Refusal requires stated grounds and, in institutional settings, documentation.

Conditional Adoption

Adoption proceeds under conditions stated in advance: what additional checks are required, which populations need added protection, what data will be monitored for distributional harm, and what threshold of evidence would trigger withdrawal. The conditions are written down and binding.

Deferred Decision

The structural analysis is incomplete. Evidence about failure rates for specific populations is unavailable, or the technology is changing faster than the evidence base. A deferral names the evidence that would resolve it and sets a timeline for revisiting the decision.

The framework treats all three paths as legitimate. The ethical failure lies in proceeding without analysis, or in treating adoption as a default that needs no justification while refusal must justify itself.

Section 4

Governance Case: AI-Based Academic Integrity Detection

AI-based academic integrity tools claim to detect AI-generated text, plagiarism, or contract cheating. Colleges and universities have adopted them widely, often before anyone has analyzed them structurally. This case treats them as a governance problem and applies all three levels of the AI literacy framework.

Conceptual literacy. These systems are probabilistic classifiers. They assign a text a score based on its statistical similarity to distributions they have learned to associate with AI-generated or copied text. They do not "know" whether a given student wrote a given text. A score of 85% means the text's statistical features resemble the training distribution at a threshold the vendor set. It does not mean an 85% chance that AI wrote the text.

Structural analysis. Research on these detectors documents the hierarchies they reproduce. Liang et al. (2023) found much higher false-positive rates (human-written text classified as AI-generated) for writers whose first language is not English, and the same pattern puts writers from non-dominant linguistic backgrounds, or writing in registers far from the training distribution, at risk. An institution that deploys these systems and accepts their output as evidence in academic integrity proceedings concentrates failure costs on student populations already at the margins. The design serves institutional interests, chiefly reducing the faculty time spent evaluating integrity cases. Students bear the costs of false accusations.

The distributional evidence

Liang et al. (2023) found that Turnitin's AI detection system flagged essays written by non-native English speakers as AI-generated at a false positive rate of up to 61.3%, compared to near-zero rates for native English speakers writing in dominant academic registers. The system performed as its training data would predict.

Operative implications. An institution that deploys AI detection tools without accounting for these failure rates makes a governance decision. That decision places multilingual students and students from non-dominant linguistic backgrounds at a consistent disadvantage in academic integrity proceedings. The three-decision framework asks what decision this structural analysis makes defensible.

Activity 1

Structural Analysis of a Digital Tool

Individual analysis · 45–60 minutes

Select one digital tool in use or under consideration in your teaching or at your institution. Examples include a learning management system's analytics function, an accessibility checker, a plagiarism detection tool, a content moderation system, or any AI-assisted communication or assessment tool. Apply the three questions of structural analysis from the AI literacy framework (Cole, 2026).

What a complete response does

Describes what kind of system the tool is and what it claims to do.
Answers each structural question with evidence.
States the ethical stakes and holds the decision for Activity 2.

Prompts marked * need at least one full sentence each to submit the elective.

Identify and describe the tool. (Name, function, who deployed it, at what scale, and what it claims to do. Start with conceptual accuracy: what kind of system is this, technically?) Question 1: What hierarchies does this system reproduce? (Consider the training data or design choices. Who or what defined the "normal" against which this system measures? Whose linguistic, cultural, or behavioral patterns are centered?) Question 2: Who bears the costs when this system fails? (Identify the populations most exposed to false positives, misclassifications, or system errors. What is at stake for them: grades, access, reputation? Who escapes these risks?) Question 3: Whose interests does the design serve? (Who commissioned this system, and what problem were they trying to solve? Whose interests shaped the design priorities, and whose interests were secondary?) Structural synthesis: What does this analysis reveal about the system's ethical stakes? (Hold the operative decision for Activity 2. Here, state what you now understand about the system that you did not understand before applying the three questions.)

Contribute to the repository

Structural analyses of tools used in specific disciplines, especially where no one has documented who bears the failure costs, build the evidence institutions need for governance decisions.

Yes, share my structural analysis with the C²TC repository. CWSP reviews each item before posting it and credits you by name and department unless you ask otherwise.

Activity 2

Governance Case and Operative Decision

Two parts · 60–75 minutes

Part A asks you to apply the operative decision framework to the AI detection governance case from Section 4. Part B asks you to make and justify an operative decision for the tool you analyzed in Activity 1.

Part A

The AI Detection Case: Your Operative Decision

The structural analysis of AI-based academic integrity detection in Section 4 established three points. The tools are probabilistic classifiers with known false-positive disparities concentrated among multilingual students. Students already at the margins bear the failure costs. The design serves the institution's interest in reducing workload, and student protection is secondary.

Given this analysis, what operative decision would you recommend for your institution or department? Select a decision path and develop the full grounds for it.

What a complete response does

Ties the case decision to features from the analysis.
States conditions or contingencies for the path chosen.
Writes your own decision as a governance statement addressed to those who decide.

Prompts marked * need at least one full sentence each to submit the elective.

Decision path. (Principled refusal, conditional adoption, or deferred decision, and the main reason you consider this path defensible.) Full grounds for your decision. (Tie the path to the structural analysis. Which features (hierarchies reproduced, failure cost distribution, design interests) ground the decision? A good grounds statement lets someone who would choose a different path follow your reasoning.) Conditions or contingencies. (If you chose conditional adoption: what conditions are binding and how would you monitor them? If you chose principled refusal: what institutional or evidentiary change would make adoption defensible? If you chose deferral: what evidence would resolve the deferral and what is your timeline?)

Part B

Your Tool: Operative Decision

Return to the tool you analyzed in Activity 1. Using the structural synthesis you developed, make an operative decision (principled refusal, conditional adoption, or deferred decision) and write the full justification as you would present it to a department or institutional governance body.

Your operative decision for the tool from Activity 1. (State the decision path and the primary structural grounds.) Full justification. (Write this as a governance statement: precise, grounded in the structural analysis, and addressed to the people who decide on deployment. What would need to be documented, monitored, or changed for this decision to be defensible?) Assignment design: How would you teach students to carry out this kind of structural ethical analysis? (Design one assignment or course activity that teaches the three structural questions as an analytical procedure. What would students analyze, and what evidence of structural analysis would you look for in their work?)

Contribute to the repository

Governance statements and assignment designs for teaching structural AI ethics, especially those grounded in a specific discipline, give C²TC colleagues models they put to use in their own courses.

Yes, share my governance statement and assignment design with the C²TC repository. CWSP reviews each item before posting it and credits you by name and department unless you ask otherwise.

Your Artifact

Build Your Structural Ethics Assignment

Design an assignment that teaches students to apply the three structural questions as an analytical procedure. Complete the fields below to generate a formatted, print-ready assignment sheet grounded in your governance work from Activity 2.

What a complete response does

Presents structural analysis as steps students follow.
Specifies the form of the operative decision students make.
Gives three to five criteria separating structural analysis from general critique.
States an AI use policy with its reasoning.

Prompts marked * need at least one full sentence each to submit the elective.

Assignment title (Name the analytical task: what will students analyze, and what procedure will they use?) Course and audience (Course number and title, and the student audience.) Assignment overview (2–4 sentences for students. Present structural analysis as an analytical procedure with steps students follow.) Task description (What tool will students analyze, how will they apply the three questions, and what will they produce: a written analysis, a governance memo, a presentation?) Operative decision component (Will students make an operative decision: principled refusal, conditional adoption, or deferral? What form will that decision take: a governance memo, a recommendation, a policy draft?) Assessment criteria (3–5 criteria that distinguish excellent structural analysis from generic critique. Look for precision, specificity, and grounding in evidence.) AI use policy (State your policy and the pedagogical reasoning.)

Reading List

Selected Sources

Cole, K. K. (2026). AI literacy is not one thing: Conceptual, structural, and operative levels for higher education governance. Pedagogies: An International Journal, forthcoming. Framework: cwspwolf.com/ai_literacy_unified.html.
Liang, W., Yuksekgonul, M., Mao, Y., Wu, E., & Zou, J. (2023). GPT detectors are biased against non-native English writers. Patterns, 4(7).
Noble, S. U. (2018). Algorithms of oppression: How search engines reinforce racism. NYU Press.
Eubanks, V. (2018). Automating inequality: How high-tech tools profile, police, and punish the poor. St. Martin's Press.
Broussard, M. (2018). Artificial unintelligence: How computers misunderstand the world. MIT Press.
Crawford, K. (2021). Atlas of AI: Power, politics, and the planetary costs of artificial intelligence. Yale University Press.
Benjamin, R. (2019). Race after technology: Abolitionist tools for the new Jim Code. Polity Press.
Andrejevic, M., & Gates, K. (2014). Big data surveillance: Introduction. Surveillance & Society, 12(2), 185–196.

Where you will use this work

Name the course and term where you will use what you built in this elective, if you know them. (Optional. CWSP sends one question near the end of that term.)

Four questions about this elective

Your answers here go to CWSP without your name or email, apart from your elective responses. CWSP reads them to revise the elective. They are optional, and they send when you submit.

The readings and activities applied to a course I teach. The instructions for each activity were clear. About how many hours did the elective take you? What one change would make this elective more useful for your teaching?

Complete Elective 2

When you have finished both activities, submit your responses. Your work is auto-saved in your browser.

Submitted. A copy of your responses comes to your NC State email within a few minutes. If it has not arrived within an hour, email kkcole2@ncsu.edu. Your responses also stay saved in this browser.

Submission problem. Your responses are saved in this browser. Try again or email kkcole2@ncsu.edu.

Return to

← Digital Communication Certificate