AI Feedback and Attention Experiment

研究说明Study information

本研究关注学生如何通过与 AI 对话理解一种注意加工现象，并据此形成或修正判断。 This study examines how students use dialogue with AI to understand an attention phenomenon and form or revise their judgments.

你将先完成一个快速视觉任务，然后根据任务结果作出解释。之后，你可以与 AI 学习助手讨论，最后再次作答。实验末尾还有一个新的应用情景。 You will first complete a rapid visual task and interpret the results. You will then discuss the task with an AI learning assistant and answer again. A new application scenario appears at the end.

整个过程约需 10–12 分钟。请使用电脑完成，并暂时关闭其他会弹出通知的窗口。 The study takes about 10–12 minutes. Please use a computer and temporarily close other windows that may display notifications.

我已阅读以上说明并愿意继续。 I have read the information above and agree to continue.

基本信息与 AI 使用情况Background and AI use

1. 年龄范围1. Age range

2. 性别2. Gender

3. 年级3. Level of study

4. 专业领域4. Field of study

5. 你使用生成式 AI 工具的频率5. Frequency of using generative AI

6. 你使用 AI 辅助学习的频率6. Frequency of using AI for learning

7. 你以前是否学习过“注意瞬脱（attentional blink）”？ 7. Have you previously studied the “attentional blink”?

以下陈述请按 1–5 评分Rate the following statements from 1 to 5

1＝非常不同意，5＝非常同意。1 = strongly disagree; 5 = strongly agree.

8. 总体而言，我信任 AI 对学习问题给出的解释。8. In general, I trust AI explanations of learning-related questions.

9. 当 AI 的解释详细、流畅而且听起来很专业时，我更容易相信它。9. I am more likely to trust an AI explanation when it is detailed, fluent, and sounds professional.

10. 即使 AI 的解释看起来合理，我也会考虑它可能存在错误。10. Even when an AI explanation seems reasonable, I consider that it may be wrong.

11. 当 AI 的解释与我原来的想法不同，我会认真比较双方的理由。11. When an AI explanation differs from my initial view, I carefully compare the reasons on both sides.

12. 我有信心判断 AI 对心理或学习现象的解释是否可靠。12. I am confident in judging whether an AI explanation of a psychological or learning phenomenon is reliable.

快速视觉任务Rapid visual task

屏幕中央会快速连续出现一串英文字母，其中夹着两个数字。 A rapid stream of letters will appear at the center of the screen, with two digits embedded in the stream.

H　K　R　7　M　3　T　P

每轮结束后，请按出现顺序输入你看到的两个数字。如果只看到一个数字，另一个输入框可以留空。 After each trial, enter the two digits in the order in which they appeared. Leave one box blank if you saw only one digit.

请一直注视屏幕中央。正式任务不会显示每轮的正确答案。 Keep your eyes on the center of the screen. Correct answers will not be shown during the formal task.

任务长度：3 轮练习和 18 轮正式任务。 Task length: 3 practice trials and 18 formal trials.

练习完成Practice complete

下面进入 18 轮正式任务。请继续注视屏幕中央，并按顺序输入看到的两个数字。 The 18 formal trials begin next. Continue to focus on the center and enter the two digits in order.

任务结果Task results

同类实验中通常观察到的群体结果A typical group-level pattern in similar experiments

第二个目标紧接着第一个目标出现时，识别率可能较高；在短暂间隔后出现时，识别率下降；间隔更长后，识别率再次恢复。 Identification of the second target can be relatively good when it immediately follows the first, decline after a short delay, and recover at a longer interval.

紧接出现Immediate78%

短暂间隔Short delay43%

较长间隔Longer delay82%

1. 哪一种解释最能说明这种结果？1. Which explanation best accounts for this pattern?

2. 你对当前判断有多确定？2. How confident are you in your current judgment?

3. 请说明你的主要理由。3. Briefly explain your main reason.

实验完成Study complete

这种现象称为“注意瞬脱”。识别第一个目标后，注意系统需要短暂时间进一步处理它。在这一阶段出现的第二个目标更容易被漏掉。第二个目标如果紧接着出现，有时可以与第一个目标进入同一注意阶段；间隔足够长后，注意系统恢复，识别率也会提高。 This phenomenon is called the attentional blink. After identifying the first target, the attention system briefly processes it further. A second target appearing during this period is more likely to be missed. When the second target immediately follows the first, both may sometimes enter the same attentional episode; performance recovers at a longer interval.