概述
It matters because licensing boards hold the clinician responsible for any AI-assisted work that touches a client. The software vendor does not carry that responsibility.
深入探討
No major counseling or psychology ethics code was written with generative AI in mind, but the existing standards still apply to it. The ACA Code of Ethics (2014) has Section H on distance counseling, technology and social media. It requires counselors to understand the technology they use, explain its benefits and risks during informed consent, and protect electronic records. The APA Ethical Principles of Psychologists and Code of Conduct requires psychologists to work within their competence, get informed consent and keep information confidential. The NASW Code of Ethics added technology standards for social workers in its 2017 revision. Professional associations have since issued AI-specific guidance, and it mostly restates these themes rather than creating new duties. Four obligations come up again and again. Competence: understand what a tool does, where it fails and how it stores data before you rely on it; Informed consent: clients should know when AI records, transcribes or drafts anything about them, and they should be able to decline; Confidentiality: in the US, entering identifiable client information into a consumer chatbot that has no business associate agreement can violate HIPAA, whatever your intent; and Accountability: a note, diagnosis or treatment plan you sign is yours, even if software drafted it. State law is changing too. In 2025 Illinois passed a law that bars AI systems from providing therapy or making therapeutic decisions on their own. It still allows AI for administrative and supporting tasks under a licensed professional's oversight. Other states have considered or passed rules on mental health chatbots. Rules differ from state to state, so check with your own board. One common misconception is that de-identifying data solves every problem. Removing names does nothing about accuracy, bias, or the way a tool's output can quietly shape your clinical judgment. Another is that a vendor's "HIPAA compliant" label moves responsibility onto the vendor. It does not.
現實世界的實施
在使用環境人工智慧抄寫員記錄會話筆記之前,顧問會確認供應商將簽署 HIPAA 業務夥伴協議。她將人工智慧錄音添加到她的知情同意書中,並允許客戶選擇退出,而不會對他們的護理產生任何影響。
心理學家將去識別化的小插圖貼到聊天機器人中,以集體討論鑑別診斷。她根據 DSM-5-TR 標準和她自己的評估檢查每個建議,而不是將其複製到記錄中。
集體實踐制定人工智慧政策。它列出了批准的工具和禁止的用途,例如在消費者聊天機器人中輸入客戶名稱,並說明誰審查人工智慧草稿以及如何記錄錯誤。
一位客戶提到在會話之間使用同伴聊天機器人。治療師不帶判斷地詢問,談論在危機中不能做什麼,並確保客戶有真正的危機熱線,例如美國的 988。
The Future of AI Ethics for Counselors and Therapists
期待更詳細的指導,而不是新的道德原則。隨著工具的變化,專業協會可能會不斷更新他們的人工智慧聲明。更多州立法機構正在討論以治療師身份出現的人工智慧的限制。保險公司和醫療事故承運人也可能開始要求實踐記錄他們如何使用人工智慧。核心問題將保持不變。臨床醫師了解該工具嗎?客戶是否知情並同意?資料受到保護嗎?有執照的人仍然對臨床決策負責嗎?現在養成這些習慣的臨床醫生會發現更容易適應接下來出現的任何規則。
付諸實踐
為什麼這很重要
- 風險與安全. 災難性和日常的人工智慧危害都取決於誰了解風險以及誰能夠採取行動。
- 更明確的決策. 民眾和專業素養決定強而有力的安全政策在政治上是否可行。
- 突破炒作. 清晰的解釋可以減少炒作、實驗室公關和模糊道德劇場的影響。
留意
- 將存在風險視為科幻小說,同時能力複合。
- 混淆了表面產品安全與高度自治下的對準。
- 只給非英語和非專業觀眾留下低品質的資源。
後續步驟
- 單獨的產品危害、誤用和失控/失調風險。
- 詢問哪些證據會改變您對時間表和嚴重性的看法。
- 比起行銷主張,更喜歡主要來源和具體評估。
- 確定一條行動路徑:職業、政策、資金或技能——而不僅僅是意識。
不斷探索
常見問題
諮商師和治療師的人工智慧道德是什麼?
AI ethics for counselors and therapists means applying duties you already have, such as competence, informed consent, confidentiality and avoiding harm, to AI note-takers, chatbots and clinical decision aids. It matters because licensing boards hold the clinician responsible for any AI-assisted work that touches a client. The software vendor does not carry that responsibility.
2014 年 ACA 道德準則的哪一部分涵蓋遠距諮詢、科技和社群媒體?
2014 年 ACA 準則的 H 部分涵蓋遠距諮詢、科技和社群媒體。它要求諮詢師了解他們使用的技術,並在知情同意期間解釋其風險。
根據美國 HIPAA 規定,治療師在將可識別的客戶資訊輸入人工智慧工具之前應該確認什麼?
無論臨床醫生的意圖如何,將可識別的客戶資訊輸入供應商未簽署業務合作協議的工具都可能違反 HIPAA。
根據指南,2025 年伊利諾伊州法律做了什麼?
The Illinois law bars AI systems from providing therapy or making therapeutic decisions on their own. It still allows administrative and supporting uses under a licensed professional.
A counselor signs an AI-drafted progress note that contains an error. Who is accountable, according to the guide?
The guide says a note, diagnosis or plan you sign belongs to you, even if software drafted it. A vendor's compliance label does not move that responsibility.
Why does the guide say de-identifying data does not solve every ethical problem?
Removing names protects identity. It does not make the output accurate or unbiased, and it does not stop the output from shaping the clinician's thinking.