The simple program people poured their hearts out to

In the mid-1960s, Joseph Weizenbaum, a computer scientist at the Massachusetts Institute of Technology, wrote a program called ELIZA. People typed messages to it on a typewriter connected to MIT's shared computer, and it typed replies. He described it in a paper published in January 1966.

ELIZA had no knowledge of the world. It looked for key words in what you typed and used simple rules to turn your sentence around. "I am unhappy" might come back as a question about why you are unhappy. When it found no key word, it fell back on a stock phrase that kept you talking.

Its best-known script, called DOCTOR, played a type of therapist who mostly reflects the patient's words back. Weizenbaum chose that role on purpose: in such a conversation one side can seem to know very little about the world and it still feels natural.

Many users still felt understood. Weizenbaum later recalled that his own secretary, who knew exactly how the program worked, asked him to leave the room so she could talk to it in private. The experience disturbed him, and in his 1976 book Computer Power and Human Reason he warned about trusting machines with human judgment.

Weizenbaum's own paper made the key point. People assume the other side of a conversation understands them, and they fill in the meaning themselves. The understanding was coming from the user, not the machine. This tendency is now called the ELIZA effect.

Modern chatbots are far more capable than ELIZA. They can hold a long conversation, remember earlier messages and write fluent, warm replies. But the basic trap is the same, and stronger. A reply that sounds caring and confident feels like understanding, even though the system is producing likely text, not sharing your feelings or checking your situation.

That matters most when people turn to a chatbot for advice, comfort or company. The friendlier it sounds, the easier it is to forget what it is.

The idea behind this caseLesson: What a chatbot really does →

1. 1Remind yourself that a warm tone is a writing style, not proof that the chatbot understands you or cares what happens to you.
2. 2For health, money, legal or emotional crises, use a chatbot to prepare questions, then talk to a qualified person.
3. 3Notice if you are telling a chatbot things you would not tell a stranger, and check its privacy settings before you do.
4. 4Keep real people in the loop. If a chatbot becomes your main confidant, treat that as a signal to reach out to someone.
5. 5When a reply feels insightful, ask yourself which parts came from the chatbot and which parts you filled in yourself.