TLDR;
This video discusses adapting assessment methods in education in response to the rise of AI. It highlights the need to shift from traditional assessments that focus on recall to more future-ready assessments that emphasise conceptual understanding, application, and durable skills. The presentation addresses concerns about AI being used for cheating, suggesting that while it's a valid concern, educators should focus on redesigning assessments to be AI-resistant or AI-enabled in meaningful ways.
- The rise of AI necessitates a re-evaluation of traditional assessment methods.
- Focus should shift towards assessments that promote conceptual understanding and durable skills.
- AI can be used as a tool to enhance learning and assessment, not just for cheating.
Introduction [0:05]
The presenter introduces the topic of adapting assessment in the age of AI, emphasising the importance of viewing AI as an opportunity rather than solely a challenge. She shares her background as a former educator and her current work with AI for Education, highlighting the need for AI literacy among educators and students. The discussion aims to rethink assessment methods to prepare students for the future.
The Seismic Shift Caused by AI [2:17]
The introduction of tools like ChatGPT has caused a seismic shift in education, creating both fear and uncertainty among teachers and students. The presenter draws parallels to previous technological advancements that were initially met with similar concerns. While these shifts have influenced teaching methods, assessment structures have remained largely unchanged. The focus of the discussion is on adapting assessment practices to be more responsive to technological advances, including AI.
The Elephant in the Room: AI for Cheating [6:20]
The presenter addresses the prevalent concern of AI being used for cheating, referencing a Stanford University study that has tracked cheating behaviours among high school students for a decade. The study found that the overall percentage of students engaging in cheating behaviours remained consistent even with the introduction of generative AI in 2023. However, there has been a notable increase in the number of students using AI tools to cheat, particularly in public and charter schools.
Common Answers That Are Not Fit for Purpose [9:28]
The presenter identifies two common reactions to AI-driven cheating that are not considered comprehensive solutions: returning to pencil-and-paper assessments and relying on AI detection software. While in-person assessments have their place, an exclusive return to them can be detrimental, especially for students with learning disabilities or those from non-native language backgrounds. AI detection software is often inaccurate and biased, particularly against non-native English speakers.
The Context of Cheating Matters [13:51]
Students cheat for various reasons, and the context behind their actions is important. The presenter provides examples of students using AI-powered tools for assistance, such as composing essays through voice-to-text, proofreading assignments, or organising thoughts into outlines. It's important to consider what skills are truly being assessed and to have transparent conversations with students about their approaches to using AI.
Moving from Traditional to AI-Resistant Assessments [16:56]
The discussion shifts to moving away from traditional assessments that are easily manipulated by AI towards more AI-resistant or AI-enabled methods. Traditional assessments like essays, research papers, and homework sheets often focus on lower-order thinking skills, are disconnected from the future of work, and are vulnerable to AI misuse. The goal is to create better measures of student learning that are future-ready.
Redefining Assessment [19:19]
The presenter argues that the definition of assessment needs to be redefined. The old definition focused on measuring a student's ability to recall and apply specific skills acquired in a classroom setting. The new definition should be an ongoing process of evaluating students' conceptual understanding and their ability to demonstrate that understanding through a variety of tasks, with an emphasis on durable skills like critical thinking, communication, and collaboration.
AI-Resistant Assessments [22:34]
The presenter shares that nothing is truly AI-resistant, but some assessments are less easily manipulated by AI. These assessments focus on the process rather than just the final product, involve multiple rounds of feedback, require self-assessment and reflection, and incorporate novel applications of knowledge. Examples include in-class discussions, debates, projects, experiments, peer feedback, group work, and oral exams.
Example: From AI Vulnerable to AI Resistant [29:11]
The presenter provides an example of transforming a traditional written essay into an AI-resistant assessment. Instead of simply assigning an essay on a topic, students participate in a class discussion and then submit an essay that summarises, reflects on, and responds to specific comments from their classmates. This approach requires active listening and critical thinking, making it harder for AI to replace the student's work.
Moving to AI-Enabled Assessments [33:30]
The discussion moves towards AI-enabled assessments, where students are allowed to use AI tools in meaningful ways to enhance their learning. The presenter outlines a continuum of AI use, ranging from AI-free assignments to assignments where AI is used to produce complex outputs. The key is to ensure that students have foundational skills in place and that AI is used with oversight and permission.
A Continuum of AI Use in Assessments [35:54]
The presenter details a continuum of AI use in assessments, starting with AI-free assignments where students conduct research and draft essays in class under supervision. Next is AI-assisted assignments, where students can use AI for brainstorming, grammar checking, and spelling, while maintaining a process journal documenting their AI use. AI-augmented assignments involve students collaborating with AI throughout the writing process, annotating their essays to explain their decision-making. Finally, AI-empowered assignments allow students to use AI to produce complex outputs, such as engaging in debates with chatbots or creating social media campaigns.
Key Takeaways [45:24]
The key takeaways from the presentation are that educators need to adapt to the changing landscape of AI, redefine traditional concepts of cheating and assessment, and embrace the creative opportunities that AI offers. Assignments that are vulnerable to AI cheating typically involve a single submitted product and occur primarily outside the classroom. By adapting assessment formats and expectations, educators can empower students to be more creative and engaged in their learning.
Next Steps [48:08]
The presenter suggests that educators start by evaluating their existing assignments and identifying those that are easily manipulated by AI. They should then brainstorm ways to make these assignments more AI-resistant or AI-enabled, working collaboratively with colleagues. The goal is to embrace AI as an opportunity to enhance learning and prepare students for the future.
Q&A [49:42]
The presenter answers questions about academic integrity, emphasising the importance of having clear guidelines in place at the school and classroom level. She reiterates that there is no fully reliable AI detector and that educators should not rely on them. She also addresses concerns about AI impacting students' critical thinking skills, emphasising the need to ensure that students have foundational skills in place and are using AI tools responsibly and ethically.