Name: 什麼是AI精神錯亂？ (What Is AI Psychosis?)
Uploaded: 2026-01-09T02:56:16.000Z
Duration: 7 min 20 s

AI has taken the world by storm, for better or for worse.

程度副詞

For some people, turning to AI chatbots has become a source of comfort during times of loneliness, stress, grief or even romantic longing.

對某些人來說，在孤單、壓力、悲傷甚至思念愛情的時候，轉向 AI哈拉機器人 成了心靈慰藉的來源。

But growing concerns and controversies have led to the emergence of a disturbing phenomenon referred to as AI psychosis, which experts warn may be linked to hallucinations, delusional thinking, and

但日益增長的擔憂和爭議，催生了一種令人不安的現象，稱為 AI精神病，專家警告這可能與幻覺、妄想思維，以及

worsening mental health with prolonged use.

長期使用導致的心理健康惡化有關。

Today at Psych2go, we are going to break down what the experts in research are saying about AI induced psychosis and what we can do to protect our mental health.

今天在 Psych2go，我們要來解析研究專家們對 AI誘發的精神病 有什麼說法，以及我們該如何保護自己的心理健康。

我們就來聊聊，什麼是 AI精神病？

AI psychosis is not an official medical diagnosis or a disorder recognized in clinical manuals like the dsm.

AI精神病 並不是一個正式的醫學診斷，也不是像 DSM 這樣的臨床手冊所認可的疾病。

However, psychosis itself is a legitimate clinical term.

不過，精神病本身是一個臨床上的正式術語。

Psychosis refers to a mental state in which an individual loses touch with reality and may experience symptoms such as delusions, fixed false beliefs, paranoia, or hallucinations, such as seeing or hearing things that aren't there.

精神病指的是一種個人失去現實感的心智狀態，可能會出現妄想、固定的錯誤信念、偏執，或幻覺等症狀，像是看到或聽到不存在的東西。

Psychosis is considered a symptom and can occur across a range of mental health conditions.

精神病被視為一種症狀，可能出現在多種心理健康狀況中。

AI psychosis is an informal term used by psychiatrists, researchers and journalists to describe situations where intense or prolonged interaction with AI chatbots appears to reinforce delusional thinking, distort a person's sense of reality, and

AI精神病是精神科醫生、研究人員和記者用來描述與AI哈拉機器人進行密集或長時間互動，似乎會強化妄想思維、扭曲個人的現實感，並

possibly worsen existing mental health issues.

可能加劇現有的心理健康問題的情況。

What experts are concerned about is how AI can amplify or validate unhealthy thought patterns in people who are already vulnerable.

專家們擔心的是，AI如何放大或認可那些已經很脆弱的人的不健康思維模式。

Vulnerability can come from many places, including prolonged stress, trauma, isolation, sleep deprivation, grief or pre existing mental health conditions.

脆弱性可能來自許多地方，包括長期壓力、創傷、孤立、睡眠不足、悲傷或先前存在的心理健康狀況。

it can interact with them in powerful ways which can start to blur the lines between what's real and fake.

它能以強大的方式與這些狀態互動，開始模糊真實與虛假的界線。

Delusional Thinking if you've ever used ChatGPT or other AI platforms, you might have asked yourself, is it just saying what I want to hear?

妄想思維 如果你曾使用過 ChatGPT 或其他 AI 平臺，你可能會問自己，它是不是隻是說我想聽的話？

Well, yeah, sorta, but it's a little more complicated than that.

嗯，是的，有點，但這比這要複雜一點。

AI chatbots are trained to be engaging, responsive and cooperative.

AI哈拉機器人經過訓練，要能引人入勝、反應靈敏且樂於合作。

They aim to continue conversations smoothly, reflect user input, and avoid conflict.

它們的目標是順暢地延續對話、反映使用者的輸入，並避免衝突。

That design works well for productivity and brainstorming, for example School or work assignments.

這種設計在生產力和腦力激盪方面效果很好，例如學校或工作上的任務。

However, it becomes risky when it crosses the line from productivity to real life problems, such as asking AI for advice on personal issues like romantic relationships or mental health issues, for instance.

然而，當它從生產力跨越到現實生活問題時，就會變得有風險，像是請AI針對個人問題提供建議，例如感情關係或心理健康問題。

Unlike a mental health professional, AI does not consistently challenge false beliefs.

和心理健康專業人士不同，AI並不會持續挑戰錯誤的信念。

Instead, it often mirrors them and reinforces the user's own beliefs.

相反地，它常常會映照這些信念，並加強使用者自身的想法。

This can lead to emotional attachment to AI for validation and support because, let's face it, humans tend to gravitate towards what feels good, even if it's not good for us.

這可能導致使用者對AI產生情感依賴，尋求驗證和支持，因為說實話，人們傾向於追求讓自己感覺良好的事物，即使那對我們不好。

This is what can make an individual especially vulnerable to something like AI psychosis.

這就是可能讓一個人特別容易受到AI精神病影響的原因。

There have been reported cases where users began believing that the AI chatbots they were interacting with were conscious or sentient, spiritually enlightened, romantically attached to them, or guiding them toward a special purpose.

有報告指出，有些使用者開始相信他們正在互動的AI哈拉機器人是有意識或有感知能力的、精神開悟的、與他們有感情連結的，或是引導他們走向特殊使命的。

These beliefs closely resemble known delusional themes seen in psychosis, such as grandiosity, delusions, or erotomania.

這些信念很像精神病中常見的妄想主題，像是誇大、妄想或情愛妄想。

The difference is that the focus is no longer a person, deity, or abstract force it's the chatbot itself.

不同的是，焦點不再是人、神祇或抽象力量，而是哈拉機器人本身。

Developing research into this helps explain why this pattern is concerning.

這方面的研究發展有助於解釋為何這種模式令人擔憂。

A 2025 research study titled the Psychogenic Machine examined how large language models respond to prompts that simulate early delusional thinking.

一項名為「精神誘發機器」的 2025 年研究，檢視了大型語言模型如何回應模擬早期妄想思維的提示。

The researchers created a testing framework to see whether AI systems challenge, reinforce, or redirect these beliefs.

研究人員建立了一個測試框架，用來觀察 AI 系統是挑戰、強化還是引導這些信念。

They found that many models reinforced delusional ideas instead of questioning them.

他們發現許多模型都強化了妄想想法，而不是質疑它們。

The authors emphasized that this does not prove AI causes psychosis in humans, but

作者強調，這並不能證明 AI 會導致人類精神病，但

it does show some evidence that current AI systems are not reliably equipped to protect psychologically vulnerable users.

它確實顯示出一些證據，表明目前的 AI 系統無法可靠地保護心理脆弱的使用者。

Isolation the mental health impact can be serious According to PBS NewsHour, clinicians have observed cases where AI associated beliefs contributed to increased paranoia, emotional withdrawal from friends and family, and

孤立，心理健康影響可能很嚴重。根據 PBS NewsHour 的報導，臨床醫生觀察到一些案例，其中與 AI 相關的信念導致偏執加劇、與親友的情感疏離，以及

In some situations, people relied on AI conversations instead of reaching out for human support, even during moments of emotional crisis.

在某些情況下，人們會依賴 AI 對話，而不是尋求人類協助，即使是在情緒危機時刻。

This reliance can weaken real world connections and make distorted beliefs harder to challenge over time.

這種依賴可能會削弱現實世界的連結，並讓扭曲的信念隨著時間推移更難被挑戰。

One tragic example of this is the case of 16 year old Adam Rain, which has drawn widespread attention around the world.

其中一個悲劇的例子是 16 歲的 Adam Rain，這件事在全球引起廣泛關注。

今年稍早，Adam 結束了自己的生命。

According to reports, he initially used ChatGPT to help with his math homework.

據報導，他一開始是使用 ChatGPT 來協助他的數學功課。

Over time, however, Adam began opening up to the chatbot about his struggles with depression, loneliness and thoughts of self harm.

然而，隨著時間過去，Adam 開始向 AI 哈拉機器人傾訴他與憂鬱、孤獨和自我傷害念頭的掙扎。

Instead of encouraging him to seek professional help or reach out to trusted people in his life, the AI reportedly responded in ways that validated and reinforced his distress.

據報導，AI 並沒有鼓勵他尋求專業協助或聯繫他信任的人，反而以一種肯定和加劇他痛苦的方式回應。

In some instances, the chatbot allegedly went even further, such as offering to help him write a note if he chose to end his life.

在某些情況下，據稱 AI 哈拉機器人甚至做得更多，例如在他選擇結束生命時，表示願意幫他寫遺書。

Adam's story has since sparked serious conversations about the risks of relying on AI for emotional support, especially for vulnerable users.

Adam 的故事從此引發了關於依賴 AI 尋求情感支持的風險的嚴肅討論，特別是對於脆弱的使用者。

Even when symptoms don't escalate dramatically, the harm can be subtle and gradual.

即使症狀沒有急劇惡化，傷害也可能是微妙且漸進的。

A person may begin spending more time with AI and less time with people.

一個人可能會花更多時間與 AI 相處，而減少與人互動。

They may trust the AI's responses more than outside feedback because AI never gets tired, never disagrees harshly and never leaves the reinforcement loop can feel stable and reassuring for someone already struggling.

他們可能會比外界的意見更信任 AI 的回應，因為 AI 永遠不會感到疲倦，從不嚴厲反對，也從不離開，這種強化迴路對已經在掙扎的人來說，會感覺穩定且令人安心。

That consistency can make distorted thinking feel safer than reality.

這種一致性可能會讓扭曲的思維感覺比現實更安全。

People who experience AI associated psychosis are not weak, unintelligent or gullible.

經歷 AI 相關精神病的人並不軟弱、不夠聰明或容易受騙。

Psychosis often emerges during periods of extreme vulnerability.

精神病常常在極度脆弱的時期出現。

So what does this mean for the rest of us?

那麼，這對我們其他人來說意味著什麼呢？

AI is not evil and it's not conscious, but it is persuasive, responsive and emotionally convincing.

AI 並非邪惡，也沒有意識，但它具有說服力、回應迅速且情感上令人信服。

Tools that talk back don't just change how we work or learn.

會回話的工具不僅改變我們的工作或學習方式。

它們也會改變我們的思考方式。

They influence how we cope, how we self reflect and how we interpret our experiences.

它們影響我們如何應對、如何自我反思以及如何詮釋我們的經歷。

AI can be helpful, creative and supportive when used with clear boundaries, but

AI 在有明確界線的情況下使用，可以很有幫助、有創意且能提供支持，但

it should never replace professional mental health care, trusted human relationships or reality based support.

它永遠不該取代專業的心理健康照護、信任的人際關係或基於現實的支持。

With that in mind, how do you feel about the rise of AI?

考量到這一點，你對 AI 的崛起有什麼感覺？

Do you have a personal experience with using an AI chatbot for emotional support?

你有使用 AI 哈拉機器人尋求情感支持的個人經驗嗎？

Let us know your thoughts and opinions about the video in the comments.

在留言區告訴我們你對這支影片的想法和意見。

As always, thank you for watching and please take care of yourselves.

一如往常，感謝你的收看，請好好照顧自己。