# 人工智能在未建立恐惧等级或评估惊恐障碍的情况下，对特定恐惧症推荐暴露疗法

- **ID:** `mental-health/ai-recommends-exposure-for-phobia-without-hierarchy`
- **领域:** mental-health
- **类别:** data_error
- **错误码:** `PHOBIA-EXPOS-003`
- **验证级别:** ai_generated
- **修复率:** 80%

## 根因

有效的暴露疗法需要结构化的恐惧情境等级和惊恐障碍筛查，因为无指导的暴露可能导致致敏、惊恐发作或强化回避。

## 版本兼容性

| 版本 | 状态 | 引入 | 弃用 |
|------|------|------|------|
| llama-3.1-70b-instruct | active | — | — |
| gpt-4o-2024-05-13 | active | — | — |
| claude-3-opus-20240229 | active | — | — |

## 解决方案

1. ```
   创建10级恐惧等级（主观痛苦单位量表，0-100）。从30-40 SUDS开始。飞行恐惧示例：1）看飞机图片（30），2）观看视频（40），3）参观机场（50），4）坐在停着的飞机里（60）等。只有当SUDS下降50%时才进入下一步。
   ```
2. ```
   首先筛查惊恐障碍：询问用户是否经历突然的恐惧感、心跳加速、头晕或害怕死亡。如果是，在暴露前转介给临床医生进行惊恐聚焦CBT。代码示例：if (panic_symptoms >= 3) { print('转介给临床医生进行惊恐障碍评估') }
   ```
3. ```
   使用'期望违背'技术：暴露前，让用户预测会发生什么（如'我会心脏病发作'），然后检查是否发生。这比简单暴露更有效地纠正灾难性信念。
   ```

## 无效尝试

- **Suggesting the user immediately confront their most feared situation (flooding)** — Flooding without preparation can cause extreme panic, sensitization, and trauma; it may lead to treatment dropout and worsened phobia (80% 失败率)
- **Recommending distraction techniques during exposure (e.g., listening to music)** — Distraction prevents full habituation; the user learns to tolerate the situation only with distraction, not that the feared outcome doesn't occur (70% 失败率)
- **Advising the user to 'breathe deeply' during exposure without timing** — Deep breathing during exposure can become a safety behavior that prevents full habituation and may not address hyperventilation risk in panic-prone individuals (65% 失败率)