Artificial intelligence: when machines are considered more benevolent than us

This column is dedicated to all those who prefer to deal with a human when it comes to negotiating a price or a service. This group considers humans to be more flexible, more empathetic, more apt to improvise in order to arrive at a fair and balanced solution. This positive view of human judgement, as opposed to the brutal coldness of artificial intelligence, is being shattered as online services expand. These are favored by a generation whose approach is more transactional – I pay, so I have an expectation – which gives a bonus to efficiency, preferably measurable.

This notion of maximizing the “return on relationship” is decisive. In an intensely competitive environment like tech, it is rare for people to meet without there being transactional anticipation: a novice will seek more experienced than him to accelerate his learning curve on a specific point; the senior will accept on the basis of an evaluation of the return on time spent (this little youngster can do something interesting, it’s worth an hour of my time). A amazing study by Aaron Garvey (University of Kentucky), TaeWoo Kim (Sydney University of Technology) and Adam Duhachek (University of Sydney) sheds new light on this issue.

The researchers presented groups with different scenarios based on the purchase of a concert ticket and an Uber ride. Same principle in both cases; guinea pig customers are offered three prices: one is cheaper than expected, the second is in line with their expectations, and a final offer is more expensive than expected. Central element of the study: the offers are alternately presented by a human named Alex, and by a robotic interface, a chatbot, which responds to the metallic name of “XT-1000”. They are then asked their opinion on the quality of the interaction, according to three criteria: is the system benevolent, selfish, fair?

“For bad news, send an AI”

Conclusion: faced with a disappointing price, it is the robot which is nevertheless judged “more benevolent”, “less selfish” and “more fair”. Conversely, if the price offered for the concert ticket or the Uber ride is, in the end, more interesting than expected, it is the opposite: the human being is then seen as more altruistic and fairer in his offer. Hence the conclusion (and title) of the study: “For bad news, send an AI, for good news, put a human”. The cognitive bias is clearly favorable to machines, considered more neutral, more subject to an objective quantification of the issues, as opposed to a biased, greedy, and potentially cunning human.

However, the conclusions of this study are unlikely to accelerate the deployment of robotic interfaces in interactions with the public due to an element terribly poorly managed by digital interfaces: the exception. Admittedly, in the case of Uber for example, when it comes to integrating in a few milliseconds elements as numerous and heterogeneous as the availability of cars in an area, traffic, weather, their changes in minutes or hours to come, the AI ​​is unbeatable. She knows how to take into account a multitude of factors and return them in the form of clear and unambiguous information (the price of a ride in VTC or a plane ticket).

The Express app

To track analysis and decryption wherever you are

Download the app

Download the app

The exception is a nightmare for automated systems

But as soon as we leave a delimited and known environment, everything becomes complicated; it is for example the banking robot, incapable of understanding the statement, by the customer, of a problem which goes beyond the narrow field of his knowledge. Exiting from a standardized environment and confronting exceptions is a nightmare for automated systems. This is the case with autonomous cars, for example: many specialists in computer science believe that we may never see, at least in the foreseeable future, fully autonomous cars (so-called level 5) on the roads or in cities. . According to them, driving is just a succession of exceptions, not too badly managed by humans, through experience, adaptation and infrastructure constraints, but which are impossible to model. In the vagueness, the advantage is still, and for a long time, to the human.


opinions

Chronic

"In Therapy"a series by Eric Toledano and Olivier Nakache.Christopher Donner

Chronic

Emmanuel Macron is now alone in the face of a democracy in crisis, and a Republic in full doubt, underlines Sylvain Fort.By Sylvain Fort

Chronic

André Markowicz (here in 2012) became known by proposing to Hubert Nyssen, owner of Actes Sud, to retranslate the work of Dostoyevsky.By Pierre Assouline

Chronic

Bees in a hive in the Navajas Valley, producing coveted honey in Europe, on March 21, 2019 in Cuba.Albert Moukheiber

We want to give thanks to the writer of this write-up for this outstanding content

Artificial intelligence: when machines are considered more benevolent than us


Explore our social media accounts and also other pages related to themhttps://www.ai-magazine.com/related-pages/