An article published by Nature [1], one of the most prestigious scientific journals of the world, recently spread a study carried out by researchers from the MIT Media Lab of the Massachusetts Institute of Technology (USA). The research involved more than 40 million people all over the world and its surprising conclusions reveal, among many other things, that ethic varies depending on culture, economy and geographical location.
By: Gabriel E. Levy B. – www.galevy.com
What a self-driving vehicle must favor: Killing an old man or a baby?
The Moral Machine is the name of the study developed by researchers from the MIT Media Lab and which was launched in 2014 [2]. The experiment aimed to create a platform similar to a game, in which users would have the option to choose between different variables about the decisions that a self-driving vehicle could assume; participants had to prioritize human life in different scenarios, on the basis of an old dilemma, named “The Trolley Problem” [3] .
The Trolley Problem
As the MIT explains in its publication, the classic trolley problem is based on the following premise:
“A runaway trolley is about to crash and kill five people. We have access to a lever that we can use to make the trolley change tracks. However, in the middle of it there is another person who would die prematurely. Should we pull the lever and kill one life to save the other five?” MIT Technology Review Article [4]
Objectives and scope of the study
From the beginning, the study aimed to help understand the collective ethical priorities in different cultures. However, as a recent article from the MIT Technology Review [5] informed, “the researchers did not anticipate that their experiment would go viral”, since four years after its launch, people from 233 countries and territories have registered around 40 million comments, turning the study into one of the largest ever made on ethics and morals.
The Internet of Things as a basis for research
The Internet of Things is just the ability so that electronic devices, of daily use and of a different nature, can keep connected to the internet to benefit one or more users, without requiring regular or permanent human intervention for such connection.
“The Internet of Things (IoT) is the network of physical objects accessed through The Internet. These objects contain embedded technology to interact with internal states or the external environment. In other words, when objects can sense and communicate, it changes how and where decisions are made, and who makes them. For example, Nest thermostats”. Ahmed Banafa professor and researcher at the World Economic Forum in Davos, Switzerland [6]
An example of the Internet of Things is when fridges, washing machines, hairdryers, vacuum cleaners, doors, alarms, gardens, cars, drones, CCTV cameras, traffic lights, tollbooths, parking lots and in general those devices of daily use are connected to the internet, in order to simplify and automatize its use through the digital management of its functions online and generally with cloud backup [7].
Inasmuch as devices become in autonomous elements, there is a potential risk that they might cause damage or accidents. This is especially valid for devices that develop speed such as drones, airplanes and of course: the mythical self-driving cars.
Although the risk of an accident is not a matter exclusively of an autonomous device, there is a subtle and yet very significant difference with regard to accidents caused by human beings and that is that these devices eventually would have the computational ability to calculate the possibility that this may happen prior to a catastrophic event. Thus, in a matter of milliseconds the computer of the car could choose whether someone must die so that others can still alive, something that, of course, will have to be previously programmed by the device’s developers [8].
The nine decisions proposed by “The Moral Machine” project
Starting from The Trolley Problem basic logic, the project proposed by the MIT researchers took up the core concept, but it was extended to nine possible situations, allowing the more than 40 million participants to contribute with their opinions and decide on the most appropriate position in each one of the cases, all of them applied to self-driving vehicles:
- Prioritize humans over pets.
- Prioritize the vehicle passengers over pedestrians.
- Prioritize the number of lives saved in total.
- Save more women than men.
- Save young people over older ones.
- Prioritize healthy people over ill people.
- Save people of high social status over others of lower level.
- Prioritize law-abiding people over criminals.
- Decide whether the vehicle should continue on its way (not acting in any way) or change directions (acting).
While these nine scenarios turned out to be complex themselves for decision-making by participants, the experiment went even further and set out to offer situations in which multiple options were combined.
“For instance, one of the scenarios posed choosing whether a self-driving car should continue straight ahead to kill three elderly pedestrians or swerve into a barricade to kill three youthful passengers.”. Publication on the study of the Nature journal [9]
Geography, culture and economy as key factors to differentiate decisions
The analysis and findings thrown by the study allowed the researchers to conclude that people’s preferences differ significantly depending on the country, but are highly correlated with culture and economy [10].
For example, people from collectivist societies like China and Japan are less likely to save young people rather than the elderly. The researchers suggest that this could be because of the greater respect that these cultures have for their elders. [11]
Countries with more individualistic cultures are more likely to save young people
Graph published in the article from the MIT Technology Review: France, Greece, Canada, United Kingdom, Australia, Norway, United States, Sweden, Germany, Estonia, Denmark, Switzerland, Holland, Israel, Finland, Singapore, Japan, South Korea, China, Taiwan. If the barometer is closer to 1, respondents consider that it’s more important to save young people; if the barometer is closer to -1, respondents prefer saving the elderly; 0 is the global average.
The development of countries, civic culture and the institutions’ strength or weakness – especially the ones of the governmental order – were equally relevant to determine the different positions that participants assumed during the study:
“Participants from less developed countries and with weaker institutions are more tolerant with reckless pedestrians rather than with pedestrians who cross adequately. And participants from countries with a high level of economic inequality show greater gaps between the treatment of people with high or low social status. Article from the Mit Technology Review [12]
Countries with more individualistic cultures tend to save more lives
Another important conclusion from the study is that a greater number of people in danger does not always become the determining factor in establishing life or death, this primary factor being only in countries whose cultures are based on the criteria of individualism, as it is the case of United Kingdom and USA, where citizens prefer, as a first criteria, saving a greater amount of lives. The researchers believe that “this is because in these countries more emphasis is placed in what each individual worth.”
Graphic published in the MIT Techno Magazine article
Graph published in the article of the MIT Technology Review Journal: France, Israel, United Kingdom, Canada, United States, Norway, Australia, Singapore, Denmark, Sweden, Germany, Switzerland, Greece, Estonia, Holland, Finland, South Korea, China, Taiwan, Japan. If the barometer is closer to 1, respondents give greater importance to save more lives; if the barometer is closer to -1, respondents consider less important to save more lives; 0 is the global average.
Country preference when saving pedestrians over the vehicle occupants
While in countries such as Japan, Norway, Singapore or Denmark citizens prefer saving pedestrians over the vehicle occupants, in countries such as China or Estonia the priority focuses on saving the vehicle occupants rather than pedestrians.
Graph published in the article of the MIT Technology Review journal: Japan, Norway, Singapore, Denmark, Finland, Switzerland, Greece, Holland, South Korea, Germany, Australia, United Kingdom, Sweden, Canada, Israel, United States, France, Taiwan, Estonia, China. If the barometer is closer to 1, respondents consider that it’s more important to save pedestrians; if the barometer is closer to -1, respondents value saving the vehicle’s passengers more; 0 is the global average.
Although the graph reflects a greater awareness on the part of the population from developed countries to privilege and defend the lives of pedestrians, it is very unlikely that, from a commercial point of view, this criteria will finally be adopted, even in these countries, since commercially it would be very complex to sell a vehicle whose configuration is designed to sacrifice the life of its occupants over the life of other citizens, leaving the issue in the regulatory or legal field, since there are in many laws and regulations of the world a clear provision towards civil responsibility, which point that the device’s owner has responsibility over the others. This is one of the challenges of civil law in view of the emergency of these new problems derived from the Internet of Things – IoT-.
A study that can provide insights for IOT regulation:
One of the major challenges that all IOT developments face, especially those that involve displacement in public space such as self-driving vehicles or autonomous drones, is that there is a network of legal and juridical vacuums that will claim a prompt regulation, since these devices governed by algorithms based on Artificial Intelligence (AI) will begin to coexist in our daily spaces in less than a decade, causing potential risks that we must anticipate.
“It is not altogether clear who must be held responsible if an artificial intelligence device causes harm (for instance, in an accident with a self-driving vehicle or because of the wrong implementation of an algorithm): the original designer, the manufacturer, the owner, the user or even the AI itself. If we apply case by case solutions, we risk to uncertainty and confusion. Lack of regulation increases the likelihood of impulsive reactions, instinctive or even fueled by public anger”. Moisés Barrio in Retina from Spain [13]
But it is yet much more complex to establish with certainty which criteria should be assumed by societies and governments in order to define the autonomous programming of these devices, even by virtue of the study conclusions carried by the MIT. In regards, the researchers in charge have emphasized that the results obtained do not aim to create a moral rule about how different countries and governments should act, they have even affirmed that, in some cases, “technologists and policy makers should ignore collective public opinion” [14].
A co-author of the study, Edmond Awad, takes as an example the comparison of social status:
“It is worrisome that people find it acceptable to save people of a higher social status rather than those of a lower social status. It is important to say: ‘Hey, we could analyze that’ instead of saying: ‘Oh, maybe we should do that’.”
“The findings must be used by companies and governments as a basis for understanding how the public would react to the ethics of different design and policies decisions.”
“We use the Trolley Problem because it offers a very good way of collecting this data, but we expect that the ethics debate will not be confined to this issue. The discussion should focus on risk analysis about who is at higher or lower risk, rather than saying who will die or will not, and also possible biases should be addressed [15]“.
In any case, what both Awad and the rest of researchers possibly expect is that these findings will be useful for civil society, technologists and governments, “to deeply analyze how these results could be translated into an ethical design and regulation for AI”.
In conclusion, the MIT study published by the Nature journal contribute very valuable analysis elements for designers, technologists, civil society, large corporations and governments, allowing them to understand the possible impact that decisions made in matters of Artificial Intelligence and autonomy of devices will have. But, above all, the study is of special relevance since it anticipates an urgent debate that involves human life and that possibly will engage the humanity attention in the coming years. This type of researches and discussions, allow with arguments and evidence, anticipate controversial decisions, which somehow will end up impacting the majority of the planet habitants in the coming years.
Photo: Marc Kleen on Unsplash.com
Graphics: MIT Magazine Publication
Links to Article references and citations:
[1] Nature Magazine article: The Moral Machine experiment
[2] Nature Magazine article: The Moral Machine experiment
[3] Verne article: From the country of Spain, on The Trolley Problem
[4] Magazine Article: MIT Technology Review on The Moral Machine Experiment
[5] Magazine Article: MIT Technology Review on The Moral Machine Experiment
[6] Article of BBVAOpenmind portal
[7] Andinalink article: The challenges of the Internet of Things
[8] Newspaper article from Economista.es, regarding the dilemma of the self-driving cars
[9] Nature Magazine article: The Moral Machine experiment
[10] Magazine Article: MIT Technology Review on The Moral Machine Experiment
[11] Magazine Article: MIT Technology Review on The Moral Machine Experiment
[12] Magazine Article: MIT Technology Review on The Moral Machine Experiment
[13] Article published by Mosises Barrios in the Spanish magazine Retina
[14] Magazine Article: MIT Technology Review on The Moral Machine Experiment
[15] Magazine Article: MIT Technology Review on The Moral Machine Experiment
Disclaimer: The published articles correspond to contextual reviews or analyses on digital transformation in the information society, duly supported by reliable and verified academic and/or journalistic sources. The publications are NOT opinion articles and therefore the information they contain does not necessarily represent Andinalink’s position, nor that of their authors or the entities with which they are formally linked, regarding the topics, persons, entities or organizations mentioned in the text.