1 of 40

From safeguarding to reimagining�Purposeful assessment redesign in the Age of Generative AI

François Cilliers

Academic Development Programme, Centre for Higher Education Development & �Department of Health Sciences Education, Faculty of Health Sciences

Cheng-Wen Huang

Centre for Innovation in Learning and Teaching, Centre for Higher Education Development

2 of 40

Please complete the�attendance register

3 of 40

Please introduce yourselves later�at your tables during discussions

4 of 40

Where do we hope to get to?

By the end of this workshop, you should be able to:

  • Articulate a new understanding of the nature of the challenge AI poses to assessment
  • Analyse learning outcomes using a structured outcome typology
  • Apply an AI-responsive framework to critically evaluate and redesign your own assessments in response to GenAI

5 of 40

The problem we think we �need to solve

6 of 40

7 of 40

… and that people are trying hard to solve

8 of 40

Kings College London

Make assessments less open to be generated by AI

Very specific topics; or no genAI use allowed. This is likely only a short-term response or in a trusted environment. Redesigning assessments will likely be needed in future.

Incorporate possibility for AI use in assessments

Identify appropriate uses of genAI tools in assessment tasks, with appropriate guidelines. This is a medium to longer-term approach, where the future adoption of genAI is being recognized.

Reconceive authentic assessments

Consider other forms or approaches to assessment where the vulnerabilities of traditional assessment to genAI is a catalyst for facilitating Assessment for Learning. This may be a medium to longer term approach, depending on your context.  

9 of 40

Levels of �AI integration in assessment

9

https://aiassessmentscale.com/

10 of 40

The “two-lane” approach

  • there is no middle ground
  • Lane 1/None: use of genAI prohibited
  • Lane 2/All: any use of genAI is permitted

11 of 40

… and critique of the “two-lane” approach

“an All-or-None approach ignores the purpose of many higher education assessments in gradually and sequentially building knowledge and skills by not allowing for middle-ground assessments where genAI use is permitted within limits and assessment security is partial. It therefore risks throwing away assessments that allow students to develop their understanding gradually through research, thinking, reflection, and revision of their ideas.”

12 of 40

3 grid�approach

3

1

2

https://www.linkedin.com/pulse/rethinking-assessments-sean-mcminn-qoevc/

13 of 40

Are we solving the right problem…�… or is the tail wagging the dog?

  • these “solutions” focus narrowly
    • on one of a set of interdependent components of a curriculum
  • responses typically reactive & piecemeal
  • risks
    • allowing random tools to drive and distort assessment (and instructional) design
    • privileging GenAI capabilities over educational intent

tool-distorted rather than purpose-driven assessment design

14 of 40

Our proposal?�Old ideas�to solve�new problems

“let’s start at �the very beginning

a very good place� to start”

15 of 40

Old ideas to solve new problems?

16 of 40

17 of 40

What then is the problem to solve?

  • we need to move on from piecemeal, reactive �assessment redesign
  • instead, we must
    • redesign learning from ground up
    • navigate an AI enabled world by beginning with the end in mind
  • how to be systematic about this?
    • Framework for Purposeful Assessment Redesign
    • designed around Outcome Review Typology�for AI-Responsive Assessment

18 of 40

Outcome review typology for �AI-responsive assessment

Educational goals

Pervasive (and expanding) generative AI capabilities

19 of 40

Educational goals

Pervasive (and expanding) generative AI capabilities

Human-centric

AI-Augmented

AI-Enabled

AI-dominant

Obsolete

requires “biological” memory or capability like judgment, ethical reasoning, empathy, cultural humility, real-world contextual understanding

traditional outcomes that need to be approached differently in an AI-enabled world; AI supports but does not replace human cognition

new or enhanced outcomes that are now possible due to AI capabilities;

AI needed to achieve learning outcomes; task not possible without it

AI performs the task independently; focus shifts from manual execution by students to critical evaluation and oversight

obsolete or unnecessary; AI performs the task better than humans without requiring human interpretation, decision-making, or verification

20 of 40

A thought experiment

20

21 of 40

MCur (Community Health)

Before genAI

  • Develop and implement a community-based intervention for tuberculosis control

In a genAI enabled world

  • No different

Learning�outcome

22 of 40

Why no change?

  • human-centred aspect of TB control requires field-based experience
  • trust-building, stigma reduction, culturally sensitive health education
  • AI cannot replace direct patient and community engagement
  • crucial for developing social accountability and real-world intervention planning skills

human-centric

23 of 40

BSc (Applied Bioinformatics)

Before genAI

  • Analyse disease trends using historical epidemiological data

In a genAI enabled world

  • Use AI-driven predictive modelling to assess future disease outbreaks

Learning�outcome

24 of 40

Why the change?

  • AI expands what students can achieve
    • predictive modelling and real-time epidemiological forecasting were not previously possible at this level
  • students work with AI-powered GIS tools to generate outbreak forecasts and develop intervention strategies
  • students learn to interpret AI-driven predictions rather than simply memorizing static epidemiological trends

AI-enabled

25 of 40

MSocSc

Before genAI

  • Independently select and apply a systematic coding approach (e.g., thematic analysis, discourse analysis) to raw textual data

In a genAI enabled world

  • Removed from the curriculum!

Learning�outcome

26 of 40

Why the change?

  • AI-driven NLP tools fully automate qualitative data extraction, coding, and analysis
    • AI tools such as NVivo AI now conduct automatic transcription and sentiment analysis
    • no need for manual coding
    • manual thematic analysis redundant
  • instead of learning manual coding, students focus on how to critically interpret AI-generated qualitative insights

obsolete

27 of 40

requires “biological” memory or capability like judgment, ethical reasoning, empathy, cultural humility, real-world contextual understanding

traditional outcomes that need to be approached differently in an AI-enabled world; AI supports but does not replace human cognition

new outcomes that are now possible due to AI capabilities

AI needed to achieve learning outcomes; task not possible without it

AI performs the task independently; focus shifts from manual execution by students to critical evaluation and oversight

obsolete or unnecessary; AI performs the task flawlessly without requiring human interpretation, decision-making, or verification

Human-centric

AI-Augmented

AI-Enabled

AI-dominant

Obsolete

Keep

Modify�or�Add new

Discard

Educational goals

Pervasive (and expanding) generative AI capabilities

28 of 40

Framework for �Purposeful Assessment Redesign

(intentionally redesign�learning opportunities)

existing LO’s categorised into �5 categories

existing

learning

outcomes

    • keep unchanged
    • modify to match category
  • add new
  • discard if obsolete

🡪 updated LO’s

ensure that revised outcomes align with broader educational goals

apply typology

take

action

review alignment

redesign

purpose driven assessment

rather than tool-distorted assessment

intentionally redesign assessment

review�equity &�social justice implications

29 of 40

Traditional outcome

(or just a

traditional artefact?)

Where

students are

Even if we

keep an outcome…

… the

way to

get there

might

change

30 of 40

Categorising existing outcomes

Small-group activity

31 of 40

10 min

Access the document here:

http://bit.ly/4h7Kgal or

There are a number of outcomes for you �to choose from.

Spread yourselves out in the document!

Choose 1 or 2 to work on. Don’t agonise – just choose!

Change the font colour so other groups know which you’ve chosen.

​

Indicate which category of the typology you think the outcome falls in.

​

If it is a category that requires modification,

propose an updated version of the outcome.

32 of 40

Redesigning assessment

Drawing on revised outcomes

33 of 40

Go back to the outcome(s) you worked on.

​

Add a new heading “Proposed assessment” below your adapted version.

​

Propose ways this outcome could be assessed

10 min

34 of 40

Some final thoughts

35 of 40

What competencies truly require �“biological memory”?

Fawns T, Schuwirth L. Rethinking the value proposition of assessment at a time of rapid development in generative artificial intelligence. Med Educ.2024;58(1):14‐16 doi:10.1111/medu.15259

​

Image generated by Meta: concept conceived by R Abraham, June 2025

36 of 40

Can we let go and fundamentally �re-imagine what outcomes could be?

37 of 40

We’re not suggesting this is easy…

https://www.linkedin.com/posts/philldawson_when-ed-gurus-say-to-redesign-assessment-activity-7326025343650123777-LNoJ/

37

!!!!!

… but it is a place to start

38 of 40

Your next steps?

Download a copy of the worksheet to plan your next steps

https://bit.ly/47oae64

39 of 40

Thanks for your participation!

Please give feedback!

There’s a link to the slides at the end of the feedback ☺

40 of 40

Thank you

40