Emotions

Do Robots need Emotions?: 

Do Robots need Emotions? Blay Whitby blayw@sussex.ac.uk

Do Robots need Emotions?: 

Do Robots need Emotions? They most certainly do not.

Structure: 

Structure My claims 5 fallacies The machine case and the human case Some thought experiments Counter-arguments

My claims: 

My claims What we commonly call emotions in human cognition are an example of a legacy system. There is absolutely no point whatsoever in including emotions or items that closely resemble emotions in robots.

Defining Emotion: 

Defining Emotion There has been a lamentable lack of effort by psychologists on this important task. Not much for my claims depends on this definition.

Defining Emotion: 

Defining Emotion There has been a lamentable lack of effort by psychologists on this important task. Not much for my claims depends on this definition. Elation, sadness, fear, and anger. A somatic correlate. Not, in essence, affect or motivation.

(1) The Gamblers’ Fallacy: 

(1) The Gamblers’ Fallacy “A random event is more likely to occur because it has not happened for a period of time; A random event is less likely to occur because it has not happened for a period of time; A random event is more likely to occur because it recently happened; A random event is less likely to occur because it recently happened.”

The Gamblers’ Fallacy: 

The Gamblers’ Fallacy Is a known defect of human reasoning The only logical fallacy to support an entire industry. ‘Winsby’ ‘The Bermuda Triangle’ and similar

(2) The Anthropocentric Fallacy: 

(2) The Anthropocentric Fallacy “Human intelligence is the paradigm case of intelligence” Like Earth-centric and Helio-centric astronomy. Now (I hope) we know better. The Turing test

The Anthropocentric Fallacy: 

The Anthropocentric Fallacy About 1.5 million animal species have so far been classified.

The Anthropocentric Fallacy: 

The Anthropocentric Fallacy About 1.5 million animal species have so far been classified. Biologists estimate that there are anything from 2-100 million more animal species alive on this planet now.

The Anthropocentric Fallacy: 

The Anthropocentric Fallacy About 1.5 million species have so far been classified. Biologists estimate that there are anything from 2-100 million more species alive on this planet now. Of these 4,260 are mammals.

The Anthropocentric Fallacy: 

The Anthropocentric Fallacy Therefore mammals represent approximately 0.2% to 0.004% of terrestrial animal species.

(3) The Anthropomorphic Fallacy: 

(3) The Anthropomorphic Fallacy “ The best (sometimes the only) way to achieve a cognitive task is the way in which (we believe) humans do it.”

The Anthropomorphic Fallacy: 

The Anthropomorphic Fallacy MYCIN Deep Blue

(4)Evolutionary Neoconservatism: 

(4)Evolutionary Neoconservatism A collection of serious fallacies about evolution Like political neocons - a dangerous breed!

Evolutionary Neoconservatism: 

Evolutionary Neoconservatism A collection of serious fallacies about evolution Like political neocons - a dangerous breed Evolution cannot optimize.

Evolutionary Neoconservatism: 

Evolutionary Neoconservatism A collection of serious fallacies about evolution Like political neocons - a dangerous breed Evolution cannot optimize. Scrapheap challenge (transl.)

Evolutionary Neoconservatism: 

Evolutionary Neoconservatism For any feature of any evolved organism all we are entitled to conclude from evolution is that so far, it has not prevented the ancestors of that organism from passing on their genes.

(5) The fallacy fallacy: 

(5) The fallacy fallacy “ We are all too erudite and philosophically astute to commit any of the previous four fallacies.”

The machine case and the human case: 

The machine case and the human case If we work hard to avoid committing any of the 5 fallacies then we are highly unlikely to want emotions (or things resembling…) in robots.

The machine case and the human case: 

The machine case and the human case If we work hard to avoid committing any of the 5 fallacies then we are highly unlikely to want emotions (or things resembling…) in robots. The whole history of computing and IT is ample demonstration of this. Do you want still more?

The Cruise Missile Thought Experiment: 

The Cruise Missile Thought Experiment Imagine a team at MIT working on the next generation of cruise missiles. These will pursue targets to new locations if necessary. A postgrad from Rosalind Picard’s group at the Media Lab offers to ‘enhance’ the missile by adding ‘The Kamikaze Emotion’ to its control apparatus. Highly-researched from documentary evidence and questionnaire surveys of ex-kamikaze pilots, this will enable the missile to ‘feel’ the same sort of dedication, purpose, and satisfaction as a human might feel in its position. What difference, if any, will this make?

The Firefighters Thought Experiment: 

The Firefighters Thought Experiment When humans are trained to handle fires, they are trained to deal with what is possibly the most dynamic, unpredictable, real-time, and error-intolerant environment there is. If emotion (choose your definition) had any function - particularly one stemming from our evolutionary history - we might expect it to be an aid to decision-making in this domain. However, firefighters are universally trained on a knowledge-based approach. Emotional influence is suppressed as far as possible by their training. Instead they are taught and tested on detailed technical knowledge of buildings, combustion, and their consequences. Decision-making is proceduralized and explicit command and control structures are put in place. Is this wrong?

Counter Arguments: 

Counter Arguments The HCI argument: “Our robots will have to live and work with us. We will find it much easier to relate to them if they have emotions.”

Counter Arguments: 

Counter Arguments The Research Argument: “You are too focused on what you call ‘useful technology’. The real reason for building robots is to reverse engineer human psychology”

Slide27: 

“ But Blay, it’s the job of philosophers to make trouble.” Prof M.A Boden ( private conversation)