Premium
This is an archive article published on February 22, 2008

Computers with remorse can be humanoid: Research

Scientists in Italy have discovered that the machines can behave even more like people if they possess a sense of ‘regret’.

.

Computers are already able to learn from their mistakes, just like humans. Now, scientists in Italy have discovered that the machines can behave even more like people if they possess a sense of ‘regret’.

According to the scientists, computer programmes that mimic human decision making actually do a better job when they incorporate the emotion of regret, British newspaper ‘The Daily Telegraph’ reported on Friday.

In their study, the researchers observed how people modify their behaviour during strategic games by looking backward to what might have been their best move, once they know what the other players’ move was.

Story continues below this ad

Dr Davide Marchiori of the University of Trento probed how this could hone the performance of neural networks, computers that are roughly modelled on the brain that are able to learn.

“Introducing an approximation of the emotion of regret — basically looking back to see how much better they could have done — allowed the models to predict human behaviour more precisely than conventional economic theories can,” Dr Marchiori was quoted as saying.

Added co-researcher Massimo Warglien of Ca’ Foscari University in Venice, “Understanding how humans learn may help to design machines or artificial agents that can better interact with humans.

“Man-machine interaction has two sides — how men adapt to machines, and how machines adapt to humans. Our work may help to design machines better adapting to human ways of learning in interaction.”

Story continues below this ad

However, he said the study has revealed that humans are more “rationalisers” than rational beings — “they learn “in retrospect” — they look backwards instead of just looking forwards.

Latest Comment
Post Comment
Read Comments
Advertisement
Advertisement
Advertisement
Advertisement