On this page, you will read about vulnerabilities in assignment requirements (in relation to GenAI) that can often be fixed. You will also find information on solutions and detailed examples. There are explanations of four specific vulnerabilities, each of which is detailed on a separate page.
By clicking on the links you can view the other pages or go back to the page on written assignment without supervision.
The more information teachers give in the assignment instructions, the easier they make the job for GenAI tools to generate an adequate response. Teachers are typically conditioned to formulate clear assignment instructions, e.g., by supplying a model answer. In the age of GenAI, less is often more. Leaving out details (e.g., the specific theory students should use in their answer) or sub-questions makes the assignment more generic. Thus, producing a high-quality answer requires more reasoning and critical reflection by students.
In particular, details concerning major, well-known theories may make an assignment vulnerable to GenAI fraud. LLMs are very good at predicting the next words in a sequence based on what they have learned. These systems are trained by analyzing patterns in large amounts of text. Thus, their output suffers from “uneven factual accuracy” (Lingard, 2023): It is considerably more likely that GenAI will excel at writing a coherent and accurate text explaining a well-known, seminal theory or concept that is often cited and appears frequently in the materials the LLM was trained on than one focused on a more specialized theory or concept that appears less frequently in the training materials – it is, to put it simply, just a numbers game, after all.
See below an example of an assignment with detailed instructions followed by suggestions for improvement. Or watch this video.
Assignment instructions
Imagine that you are an intern at an agency that has recently been awarded the contract to come up with a communication campaign for the Dutch market, promoting sustainable travel by train instead of by plane within the EU. You are put in charge of the creative team working on a digital poster for this campaign, to be placed on metro and train stations within Amsterdam. Explain how you would apply Agenda-Setting Theory and the Elaboration Likelihood Model when coming up with the design of the poster. In your answer, make sure to:
a. Explain the two theories and their mechanisms (i.e., AST and ELM);
b. Link each theory to specific elements of the poster design. Make sure to link at least two elements to each of the theories.
Explanation
As you can see in the example on the right, Chat GPT does reasonably well on part A, where an explanation of theories is asked for. It does comparatively worse on part B, not accounting for the specificities of the case (target market/audience and channel). In other words, the explanation is good, while the application is too generic. It is possible, however, to make the assignment less GenAI-friendly (see below).
| Agenda-Setting Theory (AST) is a communication theory that suggest that the media plays a significant role in influencing public opinion by determining the issues and topics that people think are important. It asserts that media has the power to set the agenda and determine the salience of various issues in the public’s mind. In the context of sustainable travel by train, AST can be applied to shape public perception and prioritize the importance of this mode of transportation over air travel. On the other hand, the Elaboration Likelihood Model (ELM) is a theory of persuasion that explains how individuals process and respond to persuasive messages. ELM posits that there are two routes to persuasion: the central route and the peripheral route. The central route involves deep cognitive processing and evaluation of arguments, while the peripheral route relies on superficial cues and heuristics. By understanding the mechanisms of ELM, the design of the poster can be tailored to engage viewers through both central and peripheral processing routes.
When designing the digital poster to promote sustainable travel by train in the Dutch market, we can apply AST and ELM to specific elements of the design:
Agenda-Setting Theory (AST):
2. Elaboration Likelihood Model (ELM):
By combining the elements of AST and ELM in the design of the digital poster, the campaign aims to influence the public agenda by emphasizing the importance of sustainable travel by train, while also engaging viewers through both central and peripheral routes of persuasion. |
| While I don’t have direct knowledge of the specific communication theories discussed in your class, I can present an approach based on two popular communication theories that are often employed in such campaigns. For this scenario, let’s consider the Diffusion of Innovations and Framing Theory.
Explanation of the appropriateness of the two theories:
Linking each theory to specific elements of the poster design: |
Revised assignment instructions
Imagine that you are an intern at an agency that has recently been awarded the contract to come up with a communication campaign for the Dutch market, promoting sustainable travel by train instead of by plane within the EU. You are put in charge of the creative team working on a digital poster for this campaign, to be placed on metro and train stations within Amsterdam. Thinking back to the class discussions, which two of the communication theories we discussed then would be the most appropriate to base the poster design on? In your answer, make sure to:
Result
The response is a much less accurate answer than the original one and would not result in a passing grade. Although one of the theories was discussed in the course, it’s not the theory that’s most appropriate to base the poster design on. The application also lacks depth and accuracy. Framing the assignment instructions in this way would result in the student needing to invest considerably more effort in designing a good prompt (i.e., input to GenAI) and editing the output to provide a correct and complete answer.
In short
To revise the assignment instructions, refrain from naming the specific theories. Instead, you can, for example, connect the question to class discussions. GenAI does not know which theories you covered in the course or what you discussed in a certain meeting. Furthermore, refrain from assignments that ask to summarize or explain well-known concepts or theories. Instead focus on application and assess knowledge with on-campus exams.

