Joe Dale
@joedale.bsky.social
UK Independent modern foreign languages & technology consultant. Former host of TES MFL forum & short-listed for NAACE Impact Award 2013 for Curriculum Support
His final advice: Keep calm, carry on marking, and don't let AI dominate you. Teachers must intervene to ensure AI is used as a pedagogic tool and collaborator, not an enemy. Here is the recording: www.facebook.com/share/v/1C8d...
Redirecting...
www.facebook.com
November 9, 2025 at 8:01 PM
His final advice: Keep calm, carry on marking, and don't let AI dominate you. Teachers must intervene to ensure AI is used as a pedagogic tool and collaborator, not an enemy. Here is the recording: www.facebook.com/share/v/1C8d...
He advocated three strategies: scaffolded writing tasks, error analysis activities to build awareness of personal weaknesses, and using AI as an interactive writing coach to encourage metacognitive thinking.
November 9, 2025 at 8:01 PM
He advocated three strategies: scaffolded writing tasks, error analysis activities to build awareness of personal weaknesses, and using AI as an interactive writing coach to encourage metacognitive thinking.
Hyland concluded that a teacher-AI partnership is the best way forward, with teachers handling higher-order feedback and AI managing surface issues such as grammar and punctuation.
November 9, 2025 at 8:01 PM
Hyland concluded that a teacher-AI partnership is the best way forward, with teachers handling higher-order feedback and AI managing surface issues such as grammar and punctuation.
Another key inequality arises because AI models reflect predominantly Western academic discourse. In future, it might be possible to ring-fence data to allow teachers to select specific English varieties they value.
November 9, 2025 at 8:01 PM
Another key inequality arises because AI models reflect predominantly Western academic discourse. In future, it might be possible to ring-fence data to allow teachers to select specific English varieties they value.
He expressed concern that the power of big technical corporations like Google and OpenAI is enormous, potentially exceeding that of many governments in their influence and reach.
November 9, 2025 at 8:01 PM
He expressed concern that the power of big technical corporations like Google and OpenAI is enormous, potentially exceeding that of many governments in their influence and reach.
Hyland addressed major ethical concerns, including data safety and ownership, noting he himself had been asked by a publisher to allow AI to use his books for modelling purposes.
November 9, 2025 at 8:01 PM
Hyland addressed major ethical concerns, including data safety and ownership, noting he himself had been asked by a publisher to allow AI to use his books for modelling purposes.
Surface-level corrections only improve the current draft unless the student actively engages and reflects on the feedback. Hyland warned that focusing too much on tech means less attention is devoted to whether feedback is received attentively.
November 9, 2025 at 8:01 PM
Surface-level corrections only improve the current draft unless the student actively engages and reflects on the feedback. Hyland warned that focusing too much on tech means less attention is devoted to whether feedback is received attentively.
Question 5: Is it Educational? Hyland questioned whether AI creates better writers rather than just better drafts. Whilst studies show errors reduce across drafts of the same text, this learning often does not transfer to other texts.
November 9, 2025 at 8:01 PM
Question 5: Is it Educational? Hyland questioned whether AI creates better writers rather than just better drafts. Whilst studies show errors reduce across drafts of the same text, this learning often does not transfer to other texts.
Others preferred human feedback because they valued face-to-face interaction and the ability to ask follow-up questions. Critically, digital literacy had a huge impact on how effectively students engaged with feedback, regardless of English proficiency.
November 9, 2025 at 8:01 PM
Others preferred human feedback because they valued face-to-face interaction and the ability to ask follow-up questions. Critically, digital literacy had a huge impact on how effectively students engaged with feedback, regardless of English proficiency.
Question 4: Is it Trusted and Valued? The valuation of AI feedback is mixed and depends heavily on digital literacy. Some students rated AI feedback highly because it was clearer, more consistent, and always available.
November 9, 2025 at 8:01 PM
Question 4: Is it Trusted and Valued? The valuation of AI feedback is mixed and depends heavily on digital literacy. Some students rated AI feedback highly because it was clearer, more consistent, and always available.
Hyland recalled a student observation that teachers' positive feedback often contains "a but after the positive", suggesting students are acutely aware of insincere or formulaic feedback, which GPT responses appeared to be.
November 9, 2025 at 8:01 PM
Hyland recalled a student observation that teachers' positive feedback often contains "a but after the positive", suggesting students are acutely aware of insincere or formulaic feedback, which GPT responses appeared to be.
Teachers use four patterns: pairing criticism with praise, hedging with modal verbs, personal expression, and questions. When asked to provide supportive feedback, ChatGPT displayed gushing positivity and vague phrases lacking insight.
November 9, 2025 at 8:01 PM
Teachers use four patterns: pairing criticism with praise, hedging with modal verbs, personal expression, and questions. When asked to provide supportive feedback, ChatGPT displayed gushing positivity and vague phrases lacking insight.
Question 3: Is it Empathetic? Hyland drew on his paper "Sugaring the Pill", which detailed how teachers mitigate critical feedback to prevent students entering a spiral of failure.
November 9, 2025 at 8:01 PM
Question 3: Is it Empathetic? Hyland drew on his paper "Sugaring the Pill", which detailed how teachers mitigate critical feedback to prevent students entering a spiral of failure.
Since ChatGPT is trained on millions of diverse internet texts, few of which are student essays, it produces author-invisible texts that fail to offer accurate feedback on rhetorical and pragmatic aspects of argument.
November 9, 2025 at 8:01 PM
Since ChatGPT is trained on millions of diverse internet texts, few of which are student essays, it produces author-invisible texts that fail to offer accurate feedback on rhetorical and pragmatic aspects of argument.
Student essays were more heavily loaded with stance features, using hedges to display uncertainty or politeness. Students also made greater efforts to engage readers using mentions, directives, and questions.
November 9, 2025 at 8:01 PM
Student essays were more heavily loaded with stance features, using hedges to display uncertainty or politeness. Students also made greater efforts to engage readers using mentions, directives, and questions.
AI essays were more formal, impersonal, and expository, containing significantly fewer interpersonal features such as hedges, boosters, attitude markers, and self-mention.
November 9, 2025 at 8:01 PM
AI essays were more formal, impersonal, and expository, containing significantly fewer interpersonal features such as hedges, boosters, attitude markers, and self-mention.
Question 2: Is it Relevant to Rhetorical Discourse? This was the most challenging area. Hyland's research showed AI-generated undergraduate texts were linguistically different from human-written ones.
November 9, 2025 at 8:01 PM
Question 2: Is it Relevant to Rhetorical Discourse? This was the most challenging area. Hyland's research showed AI-generated undergraduate texts were linguistically different from human-written ones.
In Hyland's study with Victor Zhang, comparing teachers with Pigai (used by over 13 million Chinese students), teachers addressed 16 error types whilst Pigai only picked up eight, missing many mechanical errors.
November 9, 2025 at 8:01 PM
In Hyland's study with Victor Zhang, comparing teachers with Pigai (used by over 13 million Chinese students), teachers addressed 16 error types whilst Pigai only picked up eight, missing many mechanical errors.
Research by Min and Crossway found ChatGPT feedback varied considerably even when given the same text and prompt. However, whilst ChatGPT is a tougher marker than humans, it is also more consistent in grading.
November 9, 2025 at 8:01 PM
Research by Min and Crossway found ChatGPT feedback varied considerably even when given the same text and prompt. However, whilst ChatGPT is a tougher marker than humans, it is also more consistent in grading.
Hyland structured his discussion around five crucial questions. Question 1: Is it Accurate and Consistent? AI excels at identifying grammar errors and punctuation mistakes, but consistency is a major issue.
November 9, 2025 at 8:01 PM
Hyland structured his discussion around five crucial questions. Question 1: Is it Accurate and Consistent? AI excels at identifying grammar errors and punctuation mistakes, but consistency is a major issue.
In the UK, less than a quarter of university teachers remain in the profession 10 years after they start. The hope is that Generative AI and Automated Writing Evaluation tools like Grammarly, Pigai, and ChatGPT can relieve this pressure.
November 9, 2025 at 8:01 PM
In the UK, less than a quarter of university teachers remain in the profession 10 years after they start. The hope is that Generative AI and Automated Writing Evaluation tools like Grammarly, Pigai, and ChatGPT can relieve this pressure.
Hyland addressed a fundamental problem: the growing expectation to provide more personalised, timely, and detailed feedback is overwhelming teachers, leading to fatigue and burnout.
November 9, 2025 at 8:01 PM
Hyland addressed a fundamental problem: the growing expectation to provide more personalised, timely, and detailed feedback is overwhelming teachers, leading to fatigue and burnout.
His webinar, "Feedback Fatigue: Can AI save us from ever more marking?", drew on his extensive expertise as one of the world's most respected voices in applied linguistics, the most cited linguist globally for the past five years with over 100,000 citations.
November 9, 2025 at 8:01 PM
His webinar, "Feedback Fatigue: Can AI save us from ever more marking?", drew on his extensive expertise as one of the world's most respected voices in applied linguistics, the most cited linguist globally for the past five years with over 100,000 citations.