AI and OER: Prompt tips for high-quality teaching materials

Arti­fi­cial intel­li­gence (AI) opens up new pos­si­bil­i­ties for teach­ers to design and pre­pare open edu­ca­tional resources (OER). This arti­cle presents prac­ti­cal AI tools that can reduce the work­load and increase the qual­ity of teach­ing mate­ri­als. There are spe­cial­ized AI tools for cre­at­ing var­i­ous for­mats such as texts, images and videos that can be used specif­i­cally for cre­at­ing mate­ri­als.

This is the third post in our blog series “AI in the uni­ver­sity”. The first post focused on the use of AI detec­tors the focus, fol­lowed by a legal con­sid­er­a­tion of AI and OER. We will con­clude the series with manda­tory AI skills at uni­ver­si­ties.

Deriving questions from texts: Level vs. complexity

An ini­tial area of appli­ca­tion for AI tools is the cre­ation of ques­tions for aca­d­e­mic texts that aim to delve deeper into the topic. To ensure an appro­pri­ate level, it is nec­es­sary to spec­ify require­ments for the for­mu­lated ques­tions in the prompt with regard to tar­get groups and learn­ing objec­tives. Oth­er­wise, the AI could for­mu­late very sim­ple ques­tions that do not cor­re­spond to the learn­ing objec­tives. It is there­fore advis­able to include oper­a­tors in the prompt. For exam­ple: “For­mu­late ques­tions that con­tain the fol­low­ing oper­a­tors: Ana­lyze, Eval­u­ate, Put into con­text (require­ment level 3). Put the oper­a­tor and the assigned level explic­itly in front of the for­mu­lated ques­tion.”

Illustrative case studies: Creativity vs. realism

In addi­tion to cre­at­ing ques­tions, AI tools also open up the pos­si­bil­ity of gen­er­at­ing illus­tra­tive case stud­ies. Since real-life case stud­ies in teach­ing mate­ri­als are often restricted by copy­right, AI tools offer a valu­able alter­na­tive. An AI chat­bot can cre­ate fic­tional text vignettes that address spe­cific chal­lenges. An exam­ple could be a descrip­tion of a per­son’s expe­ri­ence in a spe­cific work envi­ron­ment, which serves as a basis for dis­cus­sion.

In addi­tion to the tex­tual rep­re­sen­ta­tion, styl­ized images of the peo­ple described can be cre­ated, as they can make the mate­ri­als more vivid and pro­mote under­stand­ing. It is advis­able to request dif­fer­ent styles of pic­tures in the prompt, as pho­to­re­al­is­tic rep­re­sen­ta­tions can lead to prob­lems with regard to per­sonal rights. Alter­na­tively, styles such as comics, draw­ings or stick fig­ures can be used to make the topic visu­ally appeal­ing and legally unob­jec­tion­able.

Simulation of personas

If case stud­ies have already been cre­ated, these can be used opti­mally for an inter­ac­tive exchange with the chat­bot. The case study can be inte­grated into the prompt to ask the chat­bot to sim­u­late the peo­ple described in a sub­se­quent con­ver­sa­tion. This inter­ac­tion gives learn­ers the oppor­tu­nity to ask spe­cific ques­tions and deepen their under­stand­ing of the respec­tive chal­lenges. A pos­si­ble prompt could be: “Sim­u­late the role of a lec­turer try­ing out new inter­ac­tive meth­ods with a large group of stu­dents.”

How­ever, such sim­u­la­tions can­not and must not claim to be authen­tic. Chat­bots are based on prob­a­bil­i­ties and are not able to rep­re­sent the full com­plex­ity and indi­vid­u­al­ity of human expe­ri­ence. There­fore, they should be used as a start­ing point for dis­cus­sions and not as con­clu­sive rep­re­sen­ta­tions.

Visualization of processes: Complexity reduction vs. attention to detail

The visual rep­re­sen­ta­tion of com­plex processes is another use­ful func­tion of AI tools. They can be used to trans­late related scenes from movies into vivid image col­lages, for exam­ple, or to con­vert draw­ings doc­u­ment­ing exper­i­ment results into pub­lic domain images. How­ever, descrip­tions in the prompt should be kept as con­cise as pos­si­ble, as exces­sive detail can often lead to incon­sis­tent results that appear inac­cu­rate or con­tra­dic­tory. An exam­ple of incon­sis­tency would be a char­ac­ter wear­ing yel­low cloth­ing in the first image, but wear­ing a dif­fer­ent color in the sec­ond image, even though this was not intended.

Sim­ple graphic styles, such as stick fig­ures with color cod­ing, can help to depict a fig­ure con­sis­tently across mul­ti­ple images. An exam­ple prompt could be: “Cre­ate a pic­ture in which a green stick fig­ure helps a yel­low stick fig­ure. The envi­ron­ment should resem­ble a black and white pen­cil sketch.” How­ever, rep­re­sent­ing processes in mul­ti­ple related images often proves chal­leng­ing for AI tools, which can lead to dif­fi­cul­ties in cre­at­ing coher­ent visual rep­re­sen­ta­tions.

Caricatures and comics: copyright vs. creativity

Car­i­ca­tures and comics can pro­mote dis­cus­sions and stim­u­late cre­ative learn­ing processes. AI can help in the cre­ation of copy­right-free car­toons. How­ever, AI-gen­er­ated visual con­tent can some­times con­tain log­i­cal errors that need to be cor­rected man­u­ally in order to clearly con­vey the intended mes­sage. Post-edit­ing is often unavoid­able here, as speech and thought bub­bles can hardly be cre­ated by AI.

Charts and infographics: Clarity vs. complexity

The visu­al­iza­tion of results in the form of charts and info­graph­ics is a clear weak­ness of AI tools. Tex­tual con­tent is only imple­mented in a rudi­men­tary way, which is why it is advis­able to add it to AI-gen­er­ated images after­wards. In addi­tion, the AI often adds con­tent that was not included in the prompt, which can affect the accu­racy of fac­tu­ally designed charts.

Outlook on the inclusion of AI tools in OER

Over­all, the var­i­ous pos­si­ble uses of AI tools in teach­ing show great poten­tial to sup­port teach­ers in the cre­ation and prepa­ra­tion of OER. The gen­er­a­tion of ques­tions, the cre­ation of illus­tra­tive case stud­ies and the visual pre­sen­ta­tion of com­plex con­tent are just a few exam­ples that can enrich teach­ing. At the same time, there are rec­og­niz­able weak­nesses, for exam­ple in the rep­re­sen­ta­tion of processes in sev­eral related images or in the qual­ity of visual con­tent that has to be edited man­u­ally.

In the future, AI tools could play a cen­tral role in the fur­ther devel­op­ment of teach­ing mate­ri­als due to the increas­ing qual­ity of the results. Teach­ers could ben­e­fit from the com­bi­na­tion of dif­fer­ent AI tools, each of which spe­cial­izes in dif­fer­ent media, e.g. text gen­er­a­tion, image or video pro­duc­tion. This could increase the vari­ety and qual­ity of mate­ri­als. In addi­tion, it is becom­ing increas­ingly use­ful to refine AI-gen­er­ated con­tent after­wards, for exam­ple by using spe­cial­ized edit­ing tools to adapt the results even more specif­i­cally to one’s own needs. In this way, teach­ers can make opti­mum use of the pos­si­bil­i­ties offered by AI and at the same time com­pen­sate for the indi­vid­ual strengths and weak­nesses of the indi­vid­ual tools.

Finally, it should be borne in mind that the use of AI is energy-inten­sive and that both the train­ing data and the oper­a­tion of AI sys­tems require a high level of resources. A con­scious approach to AI tools is there­fore rec­om­mended in order to make the most of their poten­tial.

If you would like to know more about the use of AI in open edu­ca­tional mate­ri­als and are par­tic­u­larly inter­ested in the legal aspects of this, we rec­om­mend our work­shop AI and OER in use. For reg­is­tra­tion and fur­ther ques­tions, please con­tact support.twillo@tib.eu.

Supplementary materials and sources

• Pre­sen­ta­tion slides for the twillo work­shop
• MOOCs on the topic of AI at elan e.V.
• Hand­out Gen­er­a­tive AI by Vin­cent Timm
• Cre­ate AI images: These options are avail­able on CHIP.de
• AI Cam­pus: learn­ing plat­form for arti­fi­cial intel­li­gence