San Jose State Limits AI in Design Courses to Preserve Student Craft

Panel 7 - 1: Where AI Belongs: Protecting Craft and Discovery While Teaching Generative AI in Design — Ai4

Ai4

Generative AI aids visual ideation and final rendering but weakens learning during problem framing and concept selection, requiring mandatory non-AI stages to protect foundational design judgment.

AI-assisted summary of Panel 7 - 1: Where AI Belongs: Protecting Craft and Discovery While Teaching Generative AI in Design — original source. Check the original for full context.

Source published: 2026-08-14

Detailed notes

## Background Published on YouTube by Ai4 on August 14, 2026, this presentation features Leila Ensaniat, Assistant Professor of Design at San José State University. The talk outlines how generative AI can be intentionally integrated into design curricula using the Double Diamond framework without replacing student decision-making, creative voice, or tactile craft.

## TL;DR Drawing from collaborative research at San José State University, Leila Ensaniat addresses the lack of structured guidance for generative AI in design education. She presents a three-part pedagogical framework featuring four evaluative student questions, a three-tier departmental policy, and a seven-step workflow mapped onto the Double Diamond. The framework explicitly delineates where AI should be barred—such as initial problem framing and manual sketching—and where it provides value, including research synthesis, visual divergence, and final contextual rendering. Ultimately, the model ensures students maintain ownership of their creative judgment while developing critical technological literacy.

## Atmosphere & style The presentation carries a pragmatic, articulate, and reassuring academic tone that bypasses moral panic and blind tech enthusiasm in favor of empirical studio research. Ensaniat delivers her insights with warmth, clarity, and structured pedagogical purpose, creating an empathetic dialogue focused on student agency. Watching the talk feels like an inspiring and grounded workshop on modern curriculum design, culminating in an uplifting defense of physical craft, critique, and creative joy.

## Key takeaways - [00:04](https://www.youtube.com/watch?v=wRqp3ZMj-8w&t=4s) The primary challenge in design education is not deciding whether to ban AI, but addressing the lack of shared vocabulary and explicit guidance regarding how, when, and why AI should enter coursework. - [00:04](https://www.youtube.com/watch?v=wRqp3ZMj-8w&t=4s) AI adoption in design curriculum must focus on helping students leverage generative tools without losing their creative voice, craftsmanship, or critical decision-making. - [02:59](https://www.youtube.com/watch?v=wRqp3ZMj-8w&t=179s) Before using generative tools, students should evaluate four critical questions covering expected outcome, tool capabilities and limitations, output accuracy, and potential ethical harm. - [02:59](https://www.youtube.com/watch?v=wRqp3ZMj-8w&t=179s) A comprehensive pedagogical AI framework rests on three pillars: establishing a shared common language, providing structured project templates, and implementing adaptable departmental policies. - [04:16](https://www.youtube.com/watch?v=wRqp3ZMj-8w&t=256s) Applying AI to the Double Diamond framework requires preserving non-AI stages, strictly mandating manual methods during problem framing and initial hand-drawn ideation. - [04:16](https://www.youtube.com/watch?v=wRqp3ZMj-8w&t=256s) Generative AI serves best as an exploratory collaborator in background discovery, aesthetic mood boarding, rapid visual divergence, and realistic contextual rendering. - [04:16](https://www.youtube.com/watch?v=wRqp3ZMj-8w&t=256s) Concept selection and final design direction must rely strictly on human judgment, requiring students to clearly justify and articulate their rationale. - [09:32](https://www.youtube.com/watch?v=wRqp3ZMj-8w&t=572s) Generative AI delivers unequal value across design phases, excelling at early research and visual variations while offering significantly less utility during problem definition and strategic selection. - [09:32](https://www.youtube.com/watch?v=wRqp3ZMj-8w&t=572s) A three-tier policy framework—Tier 1 (No AI), Tier 2 (AI with Constraints), and Tier 3 (AI Plus)—enables faculty to calibrate AI usage to specific assignment learning outcomes. - [11:34](https://www.youtube.com/watch?v=wRqp3ZMj-8w&t=694s) Despite rapid advancements in generative technology, foundational design fundamentals—including the pencil, the physical prototype, studio critique, and creative joy—remain irreplaceable.

## Detailed walkthrough ### [00:04](https://www.youtube.com/watch?v=wRqp3ZMj-8w&t=4s) Addressing the AI Guidance Gap in Design Education

Leila Ensaniat introduces collaborative research conducted alongside her San José State University colleagues Marjan Khatibi and Dana Ragouzeos across the departments of industrial, graphic, and interaction design. The team set out to update foundational and intermediate design coursework in response to the rapid influx of generative AI tools. Early in the initiative, they discovered that the fundamental challenge was not simply determining whether students should or should not use AI. Instead, design education suffered from a complete absence of shared vocabulary and coherent guidance outlining how, when, and why AI should enter student coursework, leaving students uncertain about which tools were sanctioned and when AI shifted from supporting their learning to doing the work for them.

To establish an evidence-based foundation during the 2025–2026 academic year, the researchers executed an extensive mixed-methods research initiative across the university and design industry: - Hosted 8 collaborative workshops with faculty and students - Surveyed more than 150 students and faculty members - Interviewed 10 professional industry experts - Collected structured feedback from over 250 students - Logged more than 20 hours of structured in-class AI experimentation - Conducted over 300 hours of out-of-class tool experimentation alongside more than 300 hours of secondary literature research

At the heart of this comprehensive investigation sits a guiding question: "How can design students use AI in their creative process without losing their own voice?" Ensaniat stresses that this inquiry intentionally bypasses polarizing debates about whether AI is inherently "good or bad" or whether it should be banned outright. Instead, the initiative focuses directly on discovering where AI genuinely belongs across research, ideation, experimentation, and final delivery, while safeguarding student curiosity, craftsmanship, personal responsibility, and ultimate decision-making.

```mermaid flowchart TD accTitle: Three Pillars of the AI Design Framework subgraph Solutions["Core Framework Solutions"] L["1. Shared Common Language (Clarifying how, why, and when AI is used)"] T["2. Faculty Project Template (Scaffolding AI-augmented design briefs)"] P["3. Flexible Departmental Policy (3-tier guidelines adaptable per assignment)"] end L --> T --> P ```

[02:59](https://www.youtube.com/watch?v=wRqp3ZMj-8w&t=179s) Four Guiding Questions and the Three-Part Educational Framework

To establish thoughtful, ethical, and creative AI adoption, the research team devised a tripartite solution for design departments: creating a shared language for how and why AI enters coursework, designing structured project templates for AI-augmented assignments, and formalizing a multi-tiered department policy that faculty can customize for specific learning objectives. Surveys revealed that before touching any generative tool, students need to be trained to critically evaluate their workflow by confronting four fundamental questions:

- **Outcome:** "Will applying generative AI result in a better outcome?" - **Capabilities & Limits:** "Do I know what Gen AI tools can and can't do?" - **Trust & Verification:** "To what extent can I trust the accuracy of generative AI tools' output?" - **Harm & Ethics:** "If I use AI tools, who will they harm?"

[04:16](https://www.youtube.com/watch?v=wRqp3ZMj-8w&t=256s) Mapping Generative AI onto the Seven Steps of the Double Diamond

Ensaniat outlines how AI integrates into the traditional design process by mapping tools across seven discrete steps aligned with the classic Double Diamond framework (Discover, Define, Develop, Deliver). At each milestone, faculty and students evaluate whether AI belongs and whether its use supports or supplants critical learning goals.

```mermaid flowchart LR accTitle: Mapping AI to the 7 Design Steps subgraph Diamond1["Diamond 1: Problem Space"] S1["1. Frame Challenge NO AI (Keep own voice)"] S2["2. Discover AI Supported (Research)"] S3["3. Define Context AI Supported (Mood & feeling)"] S1 --> S2 --> S3 end subgraph Diamond2["Diamond 2: Solution Space"] S4["4. Diverge: Hand NO AI (Sketches & models)"] S5["5. Diverge: AI AI Supported (Rapid visuals)"] S6["6. Decide Direction AI Supported (Human judgment)"] S7["7. Deliver AI Supported (Storyboards & context)"] S4 --> S5 --> S6 --> S7 end Diamond1 --> Diamond2 ```

The seven-step pedagogical workflow breaks down as follows:

1. **Framing the Challenge:** Students must immerse themselves directly in the problem, observing user needs firsthand and articulating the challenge in their own words. To keep the project authentic and preserve the designer's personal perspective, the recommendation is strictly **No AI**. As illustrated by studio walls filled with physical sticky notes, this tactile step guarantees students "keep their own voice." 2. **Discover:** Involves conducting background research, market analysis, user interviews, and technical material investigations. AI functions effectively as a research partner, provided students independently verify findings through reliable secondary sources. 3. **Defining the Context:** Students establish aesthetic directions, materials, color palettes, and emotional tone by assembling mood boards and selecting three to five descriptive keywords. Generative AI is encouraged here to support conceptual and visual atmosphere building. 4. **Ideation — Diverge with Hand:** Students explore initial form factors and concepts using manual sketching, collage, mapping, and rapid physical mockups. This stage strictly mandates **No AI**, keeping thinking visible, personal, and grounded in foundational craft. 5. **Ideation — Diverge with AI:** Students use generative image tools (such as Midjourney, ChatGPT, and Vizcom) to rapidly branch out into wide-ranging, unexpected visual variations. The objective is exploration and surprise rather than generating finished artifacts. 6. **Decide and Define Direction:** Students select their strongest concept based on project constraints and design criteria. While AI can assist in synthesizing possibilities, the final decision must rely entirely on human judgment, with students articulating clear rationales for their choices. 7. **Deliver:** Students assemble final design presentations, high-fidelity contextual renderings, user scenarios, and storyboards. AI tools are permitted to assist in placing the finished concept realistically into its intended environment alongside human users.

[09:32](https://www.youtube.com/watch?v=wRqp3ZMj-8w&t=572s) Value Disparities and the Three-Tier Policy Framework

Aggregated feedback highlights a distinct pattern: generative AI does not deliver equal value across all phases of industrial and visual design. Students find generative tools immensely valuable during early discovery, visual divergence, and final delivery renderings, but report far less utility during problem definition and strategic concept selection. Ensaniat notes that this raises an important question for design educators: whether this discrepancy represents an temporary limitation of current software or stems from the inherently human, reflective nature of the design process itself.

```mermaid flowchart TD accTitle: Three-Tier AI Pedagogical Policy T1["Tier 1: No Generative AI • Strict focus on foundational craft • Manual sketching, physical modeling, raw discovery"] T2["Tier 2: AI with Constraints • Permitted only for specific, faculty-designated tasks • Secondary research, visual ideation, contextual staging • Cannot replace human critique or complete deliverables"] T3["Tier 3: AI Plus • Broad freedom for exploratory experimentation • Mandatory prompt citation & methodology disclosure • Critical evaluation and justification of all AI outputs"] ```

To provide instructors with actionable implementation tools, the team established a three-tiered assignment framework: - **Tier 1 (No Generative AI):** Students execute assignments entirely without AI assistance to hone core technical skills, manual drafting, and foundational creative problem-solving. - **Tier 2 (AI with Constraints):** AI is restricted to designated micro-tasks pre-approved by the instructor (such as expanding mood boards or conducting initial market scans). AI cannot generate final deliverables or replace studio critiques. - **Tier 3 (AI Plus):** Students are granted wide latitude to experiment with advanced generative pipelines, provided they rigorously cite their tools, critically critique raw AI outputs, and defend design trade-offs.

[11:34](https://www.youtube.com/watch?v=wRqp3ZMj-8w&t=694s) Preserving Craft, Critique, and Joy in Future Classrooms

Ensaniat concludes by emphasizing that faculty must become deliberate and precise in the language they use around generative AI, explicitly addressing its concrete risks, technical flaws, and ethical dilemmas rather than speaking in vague generalities. Because generative AI adoption in creative industries is inevitable, educators carry a direct responsibility to reformulate assignments so that every project builds durable, authentic capabilities.

Achieving this curricular transformation requires significant institutional time, deep empathy, and an active partnership with students who are eager to cultivate distinct creative identities. Reassuring educators that emerging technologies do not render classical design fundamentals obsolete, Ensaniat reminds the audience of the irreplaceable core of design education: "the pencil still matters, the prototype still matters, and the critique still matters. And if design education is going to integrate AI ethically, joy has to matter too."

## Notable quotes > Where does generative AI genuinely help the design process—and when can it get in the way of learning? — Leila Ensaniat

> [00:04](https://www.youtube.com/watch?v=wRqp3ZMj-8w&t=4s) How can design students use AI in their creative process without losing their own voice? — Leila Ensaniat

> [02:59](https://www.youtube.com/watch?v=wRqp3ZMj-8w&t=179s) Will applying generative AI result in a better outcome? — Leila Ensaniat

> [02:59](https://www.youtube.com/watch?v=wRqp3ZMj-8w&t=179s) Do I know what Gen AI tools can and can't do? — Leila Ensaniat

> [02:59](https://www.youtube.com/watch?v=wRqp3ZMj-8w&t=179s) To what extent can I trust the accuracy of generative AI tools' output? — Leila Ensaniat

> [02:59](https://www.youtube.com/watch?v=wRqp3ZMj-8w&t=179s) If I use AI tools, who will they harm? — Leila Ensaniat

> [04:16](https://www.youtube.com/watch?v=wRqp3ZMj-8w&t=256s) keep their own voice — Leila Ensaniat

> [11:34](https://www.youtube.com/watch?v=wRqp3ZMj-8w&t=694s) the pencil still matters, the prototype still matters, and the critique still matters. And if design education is going to integrate AI ethically, joy has to matter too. — Leila Ensaniat

## Mentioned Leila Ensaniat, Marjan Khatibi, Dana Ragouzeos, San José State University, Double Diamond, Midjourney, ChatGPT, Vizcom

## Video description Leila Ensaniat, Assistant Professor, San José State University, Department of Design

Where does generative AI genuinely help the design process—and when can it get in the way of learning? In this talk, I will use the Double Diamond and an industrial design example to show how AI’s value changes from research and idea generation to prototyping and presentation. I will share ways to use AI without replacing the hands-on practice, creative judgment, and decision-making students need to develop. Attendees will leave with a practical framework for setting clear expectations and four simple questions students can ask before using an AI tool. Join me for an honest, practical conversation about helping students use AI with purpose—and feel confident choosing not to use it when it does not belong.

[Source](https://www.youtube.com/watch?v=wRqp3ZMj-8w)

中文简报

圣何塞州立大学在设计课程中限制AI以保护学生手艺生成式AI有助于视觉构思和最终渲染，但在问题界定和概念挑选阶段会削弱学习效果，因此需要设置强制性的非AI阶段来保护基础设计判断力。基于圣何塞州立大学（San José State University）的合作研究，Leila Ensaniat 针对设计教育中生成式AI缺乏结构化指导的问题提出了应对方案。她展示了一个由三部分组成的教学框架，其中包括供学生自我评估的四个问题、三级院系政策，以及映射至双钻模型（Double Diamond）的七步工作流。该框架明确划定了应禁止使用AI的环节（例如最初的问题界定和手工草图绘制），以及AI能够提供价值的阶段（包括研究整合、视觉发散和最终的场景渲染）。该模型最终确保学生在培养批判性技术素养的同时，始终掌握对自己创造性判断力的主导权。圣何塞州立大学的 Leila Ensaniat 团队指出，设计教育的核心挑战并非是否禁用 AI，而是缺乏统一的规范语汇与明确指引，以界定工具是辅助学习还是替人代劳。 (0:04) 研究团队构建了涵盖通用语言、项目模板和分层政策的三支柱教学框架，并要求学生在调用生成式工具前，必须审视预期效果、工具能力边界、内容真实性与潜在伦理危害这四个核心问题。 (2:59) 在映射至双钻模型的七步设计工作流中，明确要求在“问题定义”与“手工发散构思”阶段严格禁止使用 AI，以守住学生的独立视角与基础手工技能，其余环节则可将 AI 作为辅助探索工具。 (4:16) 鉴于生成式 AI 在不同设计阶段展现的实用价值极不均衡，团队制定了从“完全禁用 AI”、“受限使用”到“深入探索（AI Plus）”的三级作业政策，以平衡核心工艺培养与前沿技术应用。 (9:32)

Reading summary

San José State University researchers Leila Ensaniat, Marjan Khatibi, and Dana Ragouzeos developed a pedagogical framework to integrate generative AI into design education without eroding student voice or foundational craft. Based on extensive mixed-methods research during the 2025–2026 academic year, the framework establishes shared terminology, workflow mappings, and a three-tiered assignment policy. Resolving the Guidance Gap Through Empirical Research (0:04) The researchers found that the central issue facing design departments was not whether to ban or adopt generative AI, but the absence of a shared vocabulary and clear guidance on how, when, and why tools should be used. Without explicit parameters, students struggled to differentiate between using AI to support learning and having it perform the core work for them. To address this, the team surveyed more than 150 students and faculty members, interviewed 10 industry experts, gathered structured feedback from over 250 students, ran eight collaborative workshops, and logged over 600 hours of experimentation and secondary research during the 2025–2026 academic year. Four Evaluative Questions for Student Workflow (2:59) To build critical discernment before interacting with generative tools, the framework trains students to interrogate four core areas: outcome, capabilities and limits, trust and verification, and harm and ethics. Specifically, students must assess whether applying generative AI yields a superior outcome, understand tool constraints, independently verify output accuracy, and consider potential harms caused by deploying the technology. These questions underpin a three-part pedagogical solution consisting of a shared language, structured assignment templates, and adaptable departmental policies. Mapping Generative AI across the Double Diamond (4:16) The framework maps generative tools across seven steps aligned with the Double Diamond model. AI is strictly prohibited in two stages: Framing the Challenge, which relies on direct user observation and physical mapping to preserve the student's authentic voice, and Diverge with Hand, which mandates manual sketching and physical mockups to ground concepts in foundational craft. Conversely, AI is permitted as a secondary research partner in the Discover phase and for atmosphere building in Defining Context. During later stages, students use tools such as Midjourney, ChatGPT, and Vizcom in Diverge with AI for rapid exploratory ideation rather than polished artifacts. The Decide and Define Direction stage restricts AI to synthesis while reserving final decisions exclusively for human judgment. In the final Deliver step, AI is permitted to assist in staging high-fidelity renderings, storyboards, and contextual environments. Assignment Tiering and Differential Tool Value (9:32) Student feedback revealed marked disparities in AI's perceived utility across the design cycle: generative tools were rated highly during early discovery, visual divergence, and final rendering, but provided significantly less utility during problem definition and strategic concept selection. Ensaniat notes that educators must determine whether this gap reflects temporary software limitations or the inherently human, reflective nature of the design process. To manage these dynamics, the researchers formulated a three-tier assignment policy. Tier 1 (No Generative AI) focuses exclusively on manual skills, technical drafting, and unassisted problem-solving. Tier 2 (AI with Constraints) allows restricted, faculty-designated micro-tasks like market scans while banning AI from completing final deliverables or critiques. Tier 3 (AI Plus) permits broad generative experimentation, subject to mandatory prompt citation, methodology disclosure, and critical defense of AI outputs. Ensaniat concludes that as industry adoption grows inevitable, design educators must replace vague generalities with precise language that confronts AI's ethical and technical flaws. While generative pipelines continue to evolve, classical fundamentals remain indispensable, underscoring that physical sketching, manual prototyping, human critique, and creative joy remain the core of design education.

精炼阅读摘要

圣何塞州立大学研究团队指出，设计教育当前的核心症结并非是否应当禁用生成式AI，而是缺乏统一的教学语汇与清晰的操作指引。研究主张通过建立分阶段的介入边界与政策架构，确保学生在借助AI提升研究与视觉效率的同时，保留自主表达、基础手艺与批判性判断力。设计教育的指引缺失与实证研究基础 (0:04) Leila Ensaniat与圣何塞州立大学工业设计、平面设计及交互设计系的合作研究者Marjan Khatibi和Dana Ragouzeos发现，随着生成式AI工具的快速涌入，师生面临的最大困扰在于缺乏明确标准，难以界定AI究竟是在辅助学习还是在代劳设计，导致学生对可用工具范围及自主性边界产生困惑。为摆脱“支持或禁止”的二元对立争论，研究团队在2025至2026学年展开了一套混合方法实证研究，探讨设计专业学生如何在创意流程中应用AI而不丧失个人风格与声音。该项调研包含8场师生协作工作坊、针对150多名师生的问卷调查、对10位行业专家的深度访谈、超过250份学生结构化反馈，以及累计超20小时的课堂实验、各超300小时的课外工具测试与二级文献研究。Ensaniat强调，研究的重点在于摸索AI在设计调研、构思、实验和最终交付等环节中的合理位置，同时坚决捍卫学生的好奇心、工匠技艺、个人责任与终极决策权。前置评估机制与三层教学政策模型 (2:59) 为引导学生在接触任何生成式工具前进行批判性审视，研究团队提炼出四项核心自省问题：应用生成式AI能否带来更优质的成果（Outcome）、自身是否充分了解该工具的能力与边界（Capabilities & Limits）、能在多大程度上信任AI输出内容的准确性并加以验证（Trust & Verification），以及使用该工具是否会对他人造成伦理伤害（Harm & Ethics）。基于上述反思机制，团队针对不同教学目标制定了弹性化的三层系级作业政策。第一阶（Tier 1：禁用AI）严格聚焦手绘草图、物理建模等基础工艺训练，排除技术代劳，强化原始问题解决能力；第二阶（Tier 2：受限使用AI）仅允许学生将AI用于教师指定的微观任务，如扩充情绪板或进行初步市场扫描，严禁直接生成最终成果或替代演播室批评；第三阶（Tier 3：深度探索AI）则赋予学生广泛的实验自由，但要求其必须完整披露提示词方法、严格核验AI输出内容并为设计取舍做出合理论证。双钻模型七步映射与AI价值分化 (4:16) 研究团队将AI工具系统化映射至传统双钻设计流程的七个步骤中，明确界定人机协作的介入点。在问题空间（Diamond 1）中，第一步“界定挑战”严禁使用AI，要求学生通过便签整理、用户观察等实体交互来确立个人视角；第二步“发现”与第三步“定义语境”则允许AI辅助进行背景研究、市场分析及情绪氛围构建，但学生必须独立验证资料准确性。进入方案空间（Diamond 2），第四步“手工发散”严格禁用AI，要求通过纸笔草图与实体模型保持思维的可视化与工匠感；第五步“AI发散”鼓励使用Midjourney、ChatGPT、Vizcom等工具快速拓展意料之外的视觉变体；第六步“确定方向”与第七步“交付”中，AI可协助方案综合与场景渲染，但最终方向抉择必须由人类设计师依靠专业逻辑自主决定。学生反馈数据进一步揭示出AI效用存在明显分化：工具在前期调研发现、视觉发散和最终渲染阶段表现出极高价值，但在问题界定与战略决策收敛阶段的作用相对薄弱。Ensaniat指出，这一差异究竟源于当前软件的技术局限，还是本质上折射出设计流程中反思与战略决策不可替代的人性特征，仍是设计教育界值得深入探讨的核心议题。 Ensaniat总结认为，随着生成式AI在创意产业的普及，教育工作者有责任以严谨具体的语言揭示技术缺陷与伦理风险，重新设计课程作业以构建持久的核心竞争力。她强调，前沿技术的出现并未淘汰设计基本功，铅笔绘图、物理原型、演播室批判以及创造过程中的真正乐趣，在AI时代依然不可替代。

Detailed notes — zh

## Background 本内容由频道 Ai4 于 2026 年 8 月 14 日发布在 YouTube 平台上，记录了圣何塞州立大学（San José State University）设计系助理教授 Leila Ensaniat 的专题演讲。视频简介指出，演讲结合经典“双钻模型”与工业设计案例，探讨生成式 AI 在设计各阶段的真实价值与局限，旨在为设计教学提供一套清晰的预期框架与自省准则，帮助学生在拥抱技术的同时不丢失核心设计技艺。 ## TL;DR 该演讲分享了针对设计教学中生成式 AI 应用展开的深入实证调研与教学框架。研究团队提出了学生在动用 AI 前应当自省的四大核心问题，并依托“双钻模型”将设计流程细分为七个步骤，明确指出框架定义与手绘发散等环节严禁依赖 AI，而在调研与渲染交付等阶段则支持其发挥效能。此外，演讲还提出了覆盖“完全禁用”、“约束使用”到“深度拓展”的三级教学政策指南（Three Tiers）。最终，主讲人强调在拥抱 AI 浪潮的同时，教育者必须坚守手绘、实体原型与设计的纯粹喜悦，守护学生不可替代的独特声音。 ## Atmosphere & style 整场演讲呈现出严谨扎实却又极具温度的学者风范。主讲人语调沉稳笃定、循循善诱，既有详实严谨的调研数据作为理性支撑，又流露出对传统设计匠心与学生成长的深切关怀。叙事节奏层层递进，从教学一线的真实痛点自然延伸至具操作性的框架方案，并在结尾升华为对设计初心的真诚呼唤，带给听众一种既务实清醒又备受鼓舞的共鸣感。 ## Key takeaways - [00:04](https://www.youtube.com/watch?v=wRqp3ZMj-8w&t=4s) 设计教学中引入生成式 AI 的核心挑战在于缺乏通用沟通语言与一致规范，导致学生难以界定工具边界与自主声音。 - [00:04](https://www.youtube.com/watch?v=wRqp3ZMj-8w&t=4s) 研究团队通过对 150 余名师生调研、250 多名学生反馈以及数百小时的教学与工具实验，构建了一套实证教学方案。 - [03:34](https://www.youtube.com/watch?v=wRqp3ZMj-8w&t=214s) 学生在引入 AI 前需自省四大问题：是否提升产出、是否知晓能力边界、产出是否可信，以及潜在的损害与替代风险。 - [04:16](https://www.youtube.com/watch?v=wRqp3ZMj-8w&t=256s) 在双钻模型的框架定义阶段应完全禁用 AI，确保设计师通过敏锐观察与共情形成个人独特声音与清晰问题陈述。 - [04:16](https://www.youtube.com/watch?v=wRqp3ZMj-8w&t=256s) 在手绘发散阶段必须保持手工草图与实体建模，以显性化思维过程并确立核心设计语言，严禁使用 AI 代劳。 - [04:16](https://www.youtube.com/watch?v=wRqp3ZMj-8w&t=256s) 在发现调研、语境定义、AI 发散构思及成果交付阶段，生成式 AI 工具能作为高效辅助手段拓展视野与加速表达。 - [04:16](https://www.youtube.com/watch?v=wRqp3ZMj-8w&t=256s) 在方案决策与定向上色环节，AI 仅可提供参考建议，最终裁量权必须牢牢掌握在设计师本人的主观判断中。 - [09:32](https://www.youtube.com/watch?v=wRqp3ZMj-8w&t=572s) AI 在调研发现与成果交付中助益明显，但在问题定义与决策环节价值有限，反映出设计思维的本质特征与工具局限。 - [09:32](https://www.youtube.com/watch?v=wRqp3ZMj-8w&t=572s) 针对教学目标设立三级 AI 规范：Tier 1 全程禁用夯实基础、Tier 2 约束使用明确边界、Tier 3 深度拓展并强化设计溯源。 - [11:34](https://www.youtube.com/watch?v=wRqp3ZMj-8w&t=694s) 设计教育必须守护技艺与纯粹的探索喜悦，铅笔、实体原型与真人批评依然是塑造未来设计师不可或缺的基石。 ## Detailed walkthrough ### [00:04](https://www.youtube.com/watch?v=wRqp3ZMj-8w&t=4s) 项目背景与核心挑战：在设计教学中寻找 AI 的定位圣何塞州立大学（San José State University）设计系助理教授 Leila Ensaniat 介绍了她与同事 Marjan Khatibi 和 Dana Ragouzeos 共同开展的研究课题——《AI 应在何处发挥作用：在设计教学中融入生成式 AI 的同时守护技艺与探索》。该项目由工业设计、平面设计与交互设计专业跨学科协作，旨在更新基础与中阶课程内容。研究团队在起步时便意识到，核心矛盾并非简单地讨论“学生该不该用 AI”，而是缺乏一套通用的沟通语言和一致的指导规范，导致学生不清楚在作业或设计中何时、为何以及如何使用 AI，也不明确哪些工具被允许、使用到何种程度才算合适，更难以界定 AI 究竟是在辅助学习还是在“代劳包揽”。为了应对这一挑战，团队在 2025 至 2026 学年开展了深入的实证调研与教学实验，试图解答核心问题：“设计系学生如何在不丧失自我声音的前提下，将 AI 融入创意流程之中？” - 组织了 8 场师生工作坊 - 调研了 150 余名师生并采访了 10 位行业专家 - 收集了 250 多名学生的反馈 - 进行了 20 多个小时的课堂实验与 300 多个小时的课外 AI 工具实验 - 完成了 300 多个小时的二次文献研究 ```mermaid flowchart TD accTitle: AI in Design Education Framework accDescr: 展现了研究背景下构建的综合教学解决方案三要素 subgraph Solutions["三大解决方案"] A["通用语言与判断准则 （学生端四大自省问题）"] B["双钻模型 AI 映射图谱 （七步设计流程指导）"] C["三级教学政策规范 （教师端作业适配框架）"] end A --> B --> C ``` --- ### [03:34](https://www.youtube.com/watch?v=wRqp3ZMj-8w&t=214s) 学生使用生成式 AI 前的四大自省问题根据调研发现，学生在将生成式 AI 引入学习和设计工作之前，需要具备提出并回答四个核心问题的能力。这些问题帮助学生从被动使用转向主动评估： 1. **结果增益**：“使用生成式 AI 是否会带来更好的设计产出？”（Will applying generative AI result in a better outcome?） 2. **能力边界**：“我是否清楚生成式 AI 工具能做什么、不能做什么？”（Do I know what GenAI tools can and can't do?） 3. **真实可信**：“我在多大程度上可以信任生成式 AI 输出内容的准确性？”（To what extent can I trust the accuracy of generative AI tools output?） 4. **潜在代价**：“如果我使用了 AI 工具，这可能会对谁造成损害或取代谁的工作？”（If I use AI tools, who will they harm?） --- ### [04:16](https://www.youtube.com/watch?v=wRqp3ZMj-8w&t=256s) 基于双钻模型的七步设计流程与 AI 介入评估研究团队依托经典的设计“双钻模型”（Double Diamond）——从问题识别、调研分析、综合洞察，到概念发散、原型制作与测试评估——梳理出设计全流程的七个细分步骤，并逐一探讨 AI 的合理介入方式。 ```mermaid flowchart LR accTitle: 7 Steps of Design Process and AI Integration accDescr: 展示了从框架定义到交付的七个设计阶段及其 AI 适配建议 subgraph Diamond1["第一钻：发现与定义"] S1["1. 框架定义 Framing the Challenge 【严禁 AI】"] S2["2. 发现调研 Discover 【支持使用】"] S3["3. 语境定义 Define Context 【支持使用】"] end subgraph Diamond2["第二钻：开发与交付"] S4["4. 手绘发散 Ideate (Hand) 【严禁 AI】"] S5["5. AI 发散 Ideate (AI) 【支持使用】"] S6["6. 决策定向 Decide Direction 【辅助建议/人做主】"] S7["7. 方案交付 Deliver 【支持使用】"] end S1 --> S2 --> S3 --> S4 --> S5 --> S6 --> S7 ``` - **第一步：框架定义（Framing the Challenge）——【建议完全不用 AI】** 设计者必须亲自理解设计对象、目标受众与核心痛点，需要敏锐观察、共情并用自己的语言清晰写下问题陈述。正如工作室现场满墙的手写便签，这一步是确保“设计作品属于你自己并拥有独特声音”的基石。 - **第二步：发现调研（Discover）——【支持使用 AI】** 收集用户、品牌、材料、技术和市场信息，并使用可靠信源交叉验证关键事实。AI 可在此阶段高效辅助信息归纳与背景探索。 - **第三步：语境与情绪定义（Define Context）——【支持使用 AI】** 确定项目的材质、色彩、形态、情绪板（Moodboard）和视觉基调，并提炼 3 至 5 个核心关键词。AI 工具在此阶段可作为有效的视觉参考辅助。 - **第四步：手绘发散构思（Ideate: Diverge with Hand）——【建议完全不用 AI】** 运用草图、拼贴、思维导图或快速手工模型展开构思，保持思维过程外显可见。这一步是确立个人独特设计语言的核心环节。 - **第五步：AI 辅助发散构思（Ideate: Diverge with AI）——【支持使用 AI】** 利用 Midjourney、ChatGPT、Nano Banana、Vizcom 等图像生成工具快速生成海量图像。重点在于探索意外可能性与广阔概念范围，而非直接产出最终成品。 - **第六步：决策与定向上色（Decide & Defining the Direction）——【辅助支持，严禁 AI 决策】** 筛选最契合问题定义与项目目标的方向。学生需阐述为何做出该选择，AI 可提供参考建议，但终审裁量权必须掌握在设计师本人的主观判断中。 - **第七步：最终方案交付（Deliver）——【支持使用 AI】** 制作清晰阐明方案、概念、材质、用户情境及故事板（Storyboard）的最终成果展示，借助 AI 工具将最终产品置于真实使用场景中进行视觉化呈现。 --- ### [09:32](https://www.youtube.com/watch?v=wRqp3ZMj-8w&t=572s) AI 在各阶段的价值差异与三级教学规范策略调研数据显示，AI 在设计各流程中的价值分布并不均衡：学生普遍认为 AI 在“调研发现（Discover）”、“方案发散（Diverge）”和“成果交付（Deliver）”中极具助益，但在“问题定义（Define）”和“概念决策（Concept/Decide）”环节的作用却相对有限。这引发了教学层面的深层思考：这种不均衡究竟是现有工具的功能局限，还是设计思维本质特性的必然体现？为了让指导原则切实落地，团队制定了一套供教师针对不同课程和作业目标灵活调配的“三级 AI 教学指南（Three Tiers）”： | 层级名称 | 规则要求与适用场景 | 核心教学目的 | |:--- |:--- |:--- | | **Tier 1: 无生成式 AI (No Generative AI)** | 全程严禁使用任何生成式 AI 工具，学生完全凭借传统与手动方式完成作业。 | 重点训练和夯实设计基础技能，培养原生设计思考。 | | **Tier 2: 约束使用 AI (AI with Constraints)** | 仅允许在教师指定的特定环节使用（如辅助二次调研、思路扩展、局部视觉探索或排版展示）；严禁用于替代作业本身或取代真人批评反馈。 | 规范工具边界，强化人际协同与批判性评估。 | | **Tier 3: 深度拓展 AI (AI Plus)** | 赋予学生高度自由去探索实验，但必须严格注明 AI 的使用细节，评估其输出质量，并深入论证每一个设计决策背后的理由。 | 培养高阶批判性思维、工具掌控力与设计责任意识。 | --- ### [11:34](https://www.youtube.com/watch?v=wRqp3ZMj-8w&t=694s) 教学反思与结语：坚守设计的本真与喜悦在总结发言中，Leila Ensaniat 强调了教学语境与责任的多维重塑：首先，**语言至关重要（Language is important）**。在讨论生成式 AI 的价值、危害与风险时，教师必须保持严谨、审慎且明确的表述。生成式 AI 的普及已势不可挡，教育者的使命是引导下一代专业人才透彻理解其长处、局限与潜在风险。其次，**审视教学任务的设计**。教育者应当借此契机重新审视现有的作业与练习，确保布置的任务能够切实赋能学生，培养他们未来所需的真本领。这无疑需要投入大量心力与时间去探索和调试。最后，**倾听学生的心声并携手共创**。学生渴望建立属于自己的技能体系与独特声音，教育界应与学生同行，共同打造他们应得的未来设计环境。Leila 动情地总结道：“铅笔依然重要，实体原型依然重要，设计批评依然重要。如果设计教育要以符合伦理的方式融入 AI，那么设计带来的纯粹喜悦（Joy）也同样不可或缺。” ## Notable quotes > [00:04](https://www.youtube.com/watch?v=wRqp3ZMj-8w&t=4s) 核心矛盾并非简单地讨论“学生该不该用 AI”，而是缺乏一套通用的沟通语言和一致的指导规范 — Leila Ensaniat > [00:04](https://www.youtube.com/watch?v=wRqp3ZMj-8w&t=4s) 设计系学生如何在不丧失自我声音的前提下，将 AI 融入创意流程之中？ — Leila Ensaniat > [03:34](https://www.youtube.com/watch?v=wRqp3ZMj-8w&t=214s) Will applying generative AI result in a better outcome? — Leila Ensaniat > [03:34](https://www.youtube.com/watch?v=wRqp3ZMj-8w&t=214s) Do I know what GenAI tools can and can't do? — Leila Ensaniat > [03:34](https://www.youtube.com/watch?v=wRqp3ZMj-8w&t=214s) To what extent can I trust the accuracy of generative AI tools output? — Leila Ensaniat > [03:34](https://www.youtube.com/watch?v=wRqp3ZMj-8w&t=214s) If I use AI tools, who will they harm? — Leila Ensaniat > [04:16](https://www.youtube.com/watch?v=wRqp3ZMj-8w&t=256s) 设计者必须亲自理解设计对象、目标受众与核心痛点，需要敏锐观察、共情并用自己的语言清晰写下问题陈述 — Leila Ensaniat > [04:16](https://www.youtube.com/watch?v=wRqp3ZMj-8w&t=256s) 重点在于探索意外可能性与广阔概念范围，而非直接产出最终成品 — Leila Ensaniat > [04:16](https://www.youtube.com/watch?v=wRqp3ZMj-8w&t=256s) AI 可提供参考建议，但终审裁量权必须掌握在设计师本人的主观判断中 — Leila Ensaniat > [11:34](https://www.youtube.com/watch?v=wRqp3ZMj-8w&t=694s) 语言至关重要（Language is important） — Leila Ensaniat > [11:34](https://www.youtube.com/watch?v=wRqp3ZMj-8w&t=694s) 铅笔依然重要，实体原型依然重要，设计批评依然重要。如果设计教育要以符合伦理的方式融入 AI，那么设计带来的纯粹喜悦（Joy）也同样不可或缺 — Leila Ensaniat ## Mentioned Leila Ensaniat, Marjan Khatibi, Dana Ragouzeos, San José State University, Double Diamond, Midjourney, ChatGPT, Nano Banana, Vizcom ## Video description Leila Ensaniat, Assistant Professor, San José State University, Department of Design Where does generative AI genuinely help the design process—and when can it get in the way of learning? In this talk, I will use the Double Diamond and an industrial design example to show how AI’s value changes from research and idea generation to prototyping and presentation. I will share ways to use AI without replacing the hands-on practice, creative judgment, and decision-making students need to develop. Attendees will leave with a practical framework for setting clear expectations and four simple questions students can ask before using an AI tool. Join me for an honest, practical conversation about helping students use AI with purpose—and feel confident choosing not to use it when it does not belong. [Source](https://www.youtube.com/watch?v=wRqp3ZMj-8w)

Source quotes

[0:04] Leila Ensaniat: 核心矛盾并非简单地讨论“学生该不该用 AI”，而是缺乏一套通用的沟通语言和一致的指导规范 [0:04] Leila Ensaniat: 设计系学生如何在不丧失自我声音的前提下，将 AI 融入创意流程之中？ [3:34] Leila Ensaniat: Will applying generative AI result in a better outcome? [3:34] Leila Ensaniat: Do I know what GenAI tools can and can't do? [3:34] Leila Ensaniat: To what extent can I trust the accuracy of generative AI tools output? [3:34] Leila Ensaniat: If I use AI tools, who will they harm? [4:16] Leila Ensaniat: 设计者必须亲自理解设计对象、目标受众与核心痛点，需要敏锐观察、共情并用自己的语言清晰写下问题陈述 [4:16] Leila Ensaniat: 重点在于探索意外可能性与广阔概念范围，而非直接产出最终成品 [4:16] Leila Ensaniat: AI 可提供参考建议，但终审裁量权必须掌握在设计师本人的主观判断中 [11:34] Leila Ensaniat: 语言至关重要（Language is important） [11:34] Leila Ensaniat: 铅笔依然重要，实体原型依然重要，设计批评依然重要。如果设计教育要以符合伦理的方式融入 AI，那么设计带来的纯粹喜悦（Joy）也同样不可或缺

Source quotes

Leila Ensaniat: Where does generative AI genuinely help the design process—and when can it get in the way of learning? [0:04] Leila Ensaniat: How can design students use AI in their creative process without losing their own voice? [2:59] Leila Ensaniat: Will applying generative AI result in a better outcome? [2:59] Leila Ensaniat: Do I know what Gen AI tools can and can't do? [2:59] Leila Ensaniat: To what extent can I trust the accuracy of generative AI tools' output? [2:59] Leila Ensaniat: If I use AI tools, who will they harm? [4:16] Leila Ensaniat: keep their own voice [11:34] Leila Ensaniat: the pencil still matters, the prototype still matters, and the critique still matters. And if design education is going to integrate AI ethically, joy has to matter too.

Original transcript

[0:04] Okay. Hello everyone. Uh I'm so happy to be here and present our research where AI belongs [0:12] protecting craft and discovery while teaching generative AI in design. I am Leila and Sonad and [0:20] here is our team Marajan Katibi and Dana Rozos my colleagues from San Jose State University. [0:28] This project is a collaborative effort across industrial graphic and interaction design [0:34] department. We collaborated to update foundational and intermediate course content. When we started [0:42] this project, we realized that the main problem was not simply whether students should or should [0:49] not use AI. The bigger issue was that there was no shared language or consistent guidance around how, [0:59] when, and why should AI be used in the student assignment or design work. AI tools were entering [1:07] design education very quickly and students were unclear about which tools were allowed [1:13] and how much AI use was appropriate. Where does AI support learning and where does it [1:20] do the work for student? So to address this problem throughout the 2025 and 2026 academic [1:30] year we ran eight workshops with students and faculty we surveyed more than 150 students [1:37] and faculty inter we interviewed 10 industry experts gathered feedback from more than 250 [1:44] students. We conducted over 20 hours of in-class experimentation and more than 300 hour hours of [1:51] outofclass experimentation with AI tools and uh we conducted more than 300 hours of secondary [1:59] research. So here is the question at the center of this work. How can design students use AI in [2:10] their creative process without losing their own voice? Notice that this question does not ask [2:18] whether it's a it's whether AI is good or bad or whether it should allow or not. So instead it asks [2:27] where does AI blank. That is what I would like to focus on where AI belongs in the design process. [2:36] So in another words, we explored practical ways to use generative AI for research, idea generation, [2:43] experimentation, and presentation while keeping curiosity, craft, responsibility, and final design [2:50] decisions in the in the hands of designs. So uh what is our solution? We realized that what we [2:59] need is a shared framework for thoughtful, ethical and creative use of AI in design education and [3:09] uh that includes first a common language for how, why, and when AI enters a student work. Second, [3:19] a template for AI augmented project for faculty. And finally, department policy or guideline for [3:28] assignments that faculty can adopt for each project. So, I'm going to explain all these [3:34] three solutions for you in details. Based on the surveys, we find out students should be equipped [3:42] to ask and answer the following questions when considering the use of generative AI tools in [3:49] their education and in their work. Will applying generative AI result in a better outcome? Do I [3:57] know what Gen AI tools can and can't do? To what extent can I trust the accuracy the accuracy of [4:06] generative AI? AI tools output and if I use AI tools who will they hire. So before we go [4:16] further let's review this like traditional design process. Many of you might be familiar with this [4:23] double diamond diagram. Uh this is begins with identifying the problem then conduct the research [4:30] analyze and synthesize our finding and then on the second uh diamond developing the ideas [4:37] render prototype the concept and finally test and evaluate it. So what we did the question is [4:45] again does AI belong at this step? We identified seven steps in the design process and mapped the [4:54] AI onto the double diamond. At each step, we ask one simple question. Now, let's go over all these [5:03] seven steps together. The first step is about framing the challenge. Before anything else, [5:11] we should understand the problem. Understand what you are designing, who you are designing for, [5:16] who is your target, what do you want to solve? And uh to make this pro project successful and [5:23] understanding the problem, you really need to feel it. You need to observe the problem and you need [5:29] to write a paragraph explaining the problem with your own words. Does AI belong here? We recommend [5:38] no AI at this stage. Why no AI here? Because this is the step that keeps the project yours. [5:46] Designers like to have their own voice. And what you're looking at is exactly that done by uh hand [5:55] done by hand, a wall of sticky notes from one of our studio sessions. So based on the survey, [6:03] I have two uh like a gray box and the orange box. The gray box shows how do students find [6:08] AI useful at this stage at this step and the orange box shows what do faculty recommend. [6:17] Next step is to discover. At this step, we need to gather research, understand more about the user, [6:25] brand, material, technology, market and anything related to our topic and we need to verify the [6:33] important information is reliable sources. Does AI blank here? we can use AI to support this part [6:43] of the process. And as you can see some of the tools that our students uh used. The next step is [6:50] defining the context, the mood, material, color, form and overall feeling of your project. Students [6:59] need sometimes to create a mood board by or like a visual references for their project. they need [7:04] to uh choose three to five keywords to describe those images. Again, does AI blank here? Yes, [7:12] we can use AI to support this part of the process. And the next step is ideation. So, uh we um we [7:22] have two step of the ideation ideation with hand. So, we like to diverge with hand. Use sketches, [7:29] collage, mapping or quick models to develop your own ideas and to keep your own thinking visible [7:36] and for sure no AI at this stage because this is the step that keeps the project yours and to keep [7:43] your own voice. Next step, diverge with AI. We use generative image tools to generate many images [7:51] quickly. The goal is to generate a wide range of concepts and unexpected possibilities not a final [7:59] result. So we use all like tools like midjourney chpt nano banana wiscom to generate all different [8:07] possibilities and definitely is AI is supporting this step to create and surprise and see all [8:14] possibilities. Next step is decide and defining the direction. Choose the strongest directions [8:22] that best fit the problem you want to solve and the project goals. The final choice should come [8:30] from your own judgment. Students need to explain why they they select this direction. Does AI [8:37] belong here? We suggested yes AI can support you uh but it should not make the final decision for [8:45] you. And uh and the last step is deliver. Create a final presentation that clearly explains the [8:56] solution, the concept, user material and the context about your topic. So students might [9:04] use AI and because they want to create a final visuals that show all these and students might [9:12] uh need to create user scenario a story board and again does AI blind here? Yes, they can use AI to [9:22] explore all like uh final solutions and put their final product in the context along with the user. [9:32] So here is the summary of all the tools that an industrial design student might use during the [9:37] design process. And here in this chart we can see AI doesn't yet add the same value at every stage. [9:50] Student find it helpful with discover diverge and delivering the work but not so much when [9:58] you want to define your problem and your concept. Maybe there is like an a question [10:04] here. Is this a gap in the available tools or is it part of the nature of design process? [10:14] So uh to make our guidelines practical, we developed a faculty uh three guidelines for [10:21] use of AI in students project that professors can apply for each project or assignment. Uh [10:30] those trees basically three tiers include the first one no generative AI students complete [10:37] their work without generative AI so they can develop their foundational skills. The second [10:44] tier is AI with constraints. AI is allowed for specific tasks selected by the professor. It [10:53] may support secondary research, ideal expansion, visual exploration or presentation, but it cannot [11:01] complete the assignment or replace human feedback and critique. And the third one is AI plus. This [11:11] gives students more freedom to experiment but they must site AI use. They must evaluate [11:19] the output and explain their decision. So the purpose of the policy is to give students clear [11:27] expectations while allowing faculty to adopt the guidelines to the learning goals of the project. [11:34] And uh the conclusion here includes uh [11:41] language is important. How we talk about genai, its value, its harm, its risk. We [11:49] need to be intentional and be very specific in our language. Genai use inevit inevitable. We [11:57] just need to educate the next generation of professionals to truly know its strengths. [12:03] weakness and the risks involved. As educators, this is an opportunity for us to look at our [12:11] assignments and exercises and ensure that the work we are giving student is going to provide [12:17] the value and the skill development they need. And yes, this takes a lot a lot of time to do [12:24] right. And the last one which I think is very important. Our students want to develop their [12:31] own skills and unique voice. We should be listening to them and working alongside [12:38] them when we develop new ways of teaching and working to create the future that they deserve. [12:45] And uh to close this, I would say the pencil still matters, the prototype still matters [12:53] and the critic still matters. And if design education is going to integrate AI ethically, [13:01] joy has matter too. Thank you and please uh reach out to me if you have any question. Thank you.