When I was young (and indeed now), I read a lot of Isaac Asimov’s writing.
One of the things which interested Asimov was robotics (a word he more or less invented before computers, as we know them, existed).
The Three Laws of Robotics
Asimov’s robots had ‘positronic brains’ (another made-up word) and he invented the three Laws of Robotics which governed their behaviour. These were:
1) A robot may not injure a human being or, through inaction, allow a human being to come to harm;
2) A robot must obey the orders given it by human beings except where such orders would conflict with the First Law; and
3) A robot must protect its own existence as long as such protection does not conflict with the First or Second Law.
Many of his early robot stories, written in the 1940s and 1950s, were actually essays working out how these laws would operate in real life. They mostly featured a ‘robot psychologist’ called Susan Calvin, who analysed how the three laws operated in the situations in the stories.
The three laws have been very influential in science fiction, and have been read my many scientists and computer scientists. It is probable that they have also influenced thought on the ethics of artificial intelligence.
The Liar
In his early collected stories, ‘I Robot’, there is a story called ‘Liar’ written in 1941.
In this story, one of the robots, called Herbie, becomes telepathic due to an unknown fault in its manufacturing process.
As he can read minds, the first law makes Herbie give answers to questions which will reflect what the questioner wants. Particularly if he thinks the correct answer will upset the questioner.
So, Susan Calvin is told that one of her colleagues is in love with her, and another employee is told that his boss has retired and nominated him as his successor. Both untrue. When they discover the truth, they are not pleased.
Susan destroys Herbie by telling it that
- if it withheld information that her colleagues wanted, that would hurt them
- but that providing the information would hurt them too.
Herbie’s poistronic brain cannot handle what Susan described as an ‘insoluble dilemma’. It becomes insane and collapses.
Susan’s last word to the robot is ‘Liar’.
Is AI a liar?
I have been thinking a lot about this story recently, while reading about our current AI and the phenomenon of hallucination. Which is when AI makes things up.
This has become a particular problem with law and litigation, with AI often providing false information. Sometimes this is then used in court proceedings, where the false information is inevitably found out. For example
- In the 2025 case of Hamad Al‑Haroun v Qatar National Bank QPSC & QNB Capital LLC, the claimants made 45 case-law citations, 18 of which turned out to be fictitious. Quotes in many of the others were also bogus. The claimant and his solicitor admitted having used an AI tool for research.
- Also in 2025, in the case of Ayinde, R v The London Borough of Haringey, brought by Haringey Law Centre challenging the London borough of Haringey over its alleged failure to provide its client with temporary accommodation, its lawyer cited phantom case law five times. He admitted having used AI-generated research without verifying it first.
- In 2023, two New York lawyers were fined $5000 in 2023 in an aviation injury claim, Mata v. Avianca, Inc., after submitting court documents containing six non-existent cases generated by ChatGPT.
These are just a few of the many cases being reported where claimants and their lawyers are misled by false information provided by Ai tools. No doubt there are many more unreported cases.
Misleading AI information
There is also the problem of people, for example, tenants, looking things up on ChatGPT and then assuming that the answers they get are correct. Because, if (they think) the information has been provided by a computer, it must be right!
They then get annoyed when people such as lawyers, landlords and letting agents try to explain to them that the information they have been given is wrong.
Some people may even think that their lawyers, landlord or letting agent, are lying to them. Particularly those with a ‘conspiracy theory’ mindset.
A warning
ChatGPT may be useful (I use it myself sometimes) but you have to be really careful with it. In particular, be wary about relying on it for legal advice.
There have been many cases when its advice has been proven to be wrong, and the outcome for those relying on it has not been good.
After all, these machines get their information from scanning the internet, and a lot of information on the internet is wrong. As they say – garbage in, garbage out.
And finally
Many computer scientists creating our AI today will have read Asimov and have been influenced by his three Laws of Robotics.
These may have seeped into the algorithms and systems that they create.
Like Herbie the Robot in Liar, the AI tools may be influenced to fabricate false information to please you. They won’t be reading your mind like Herbie did, but no doubt they will be able to ascertain your feelings from the way you express your questions in the chat box. And maybe also from information about you on the net.
AI is still in its early stages. Perhaps we should look to Asimov’s later books to get an idea of how it will develop in future years?
Later
For a more serious post on this topic see this post from the Nearly Legal blog.
Leave a Reply