How Artificial Intelligence And Robots Simplify Our Lives - Research, Society

Bounty hunters and new algorithms

The UK division of Unilever, the global consumer goods giant, has used artificial intelligence for the first time to assess candidates for hiring.

The candidate recorded video, answering standard questions, and a machine algorithm analyzed speech, the tone in which the words were spoken, and facial expressions. As a result, hundreds of characteristics were created, which were then compared with data from candidates who were hired in the past - and they proved to be good specialists.

The American company HireVue, the developer of this application, believes that this approach is much better than the usual study of a resume. So, just by speech, the algorithm looks at 350 characteristics of the language, for example, the use of passive or active voice, the use of pronouns, vocabulary and sentence length, profession-specific terms. No human being can do that.

Such an algorithm can be used not only for sorting candidates, but also for “headhunting”. Then, if your video gets into some databases, they will find you wherever you are and make a job offer that you cannot refuse. And it will be perfect for both you and the employer

Would you like to be assessed by such an algorithm - or a person with unknown to you, but obligatory prejudices and prejudices (everyone has them), with the addition of other unpredictable and impossible to account for factors?

Thus, a person while reading a resume can be hungry or full, tired or rested. Stressful after commuting or on the first day after vacation. At the same time, the sun is shining outside the window or everything is covered with black clouds.

The algorithm in comparison with a person looks like the goddess of justice Themis - impartial and objective

But there are people who can be embarrassed at the sight of a camera, while being excellent specialists. And we know that some very talented people are different from the "norm." In addition, previous successful candidates are not always a role model in our time, when every day presents new and unexpected challenges.

But there is a more important problem, which has already been encountered by developers in other areas of application of the algorithm: on what or on whom such algorithms are trained. If the samples were taken from college graduates, whites, from families of the middle class of the east coast of the United States, then the algorithm will calculate something else that unites such people and that does not lie on the surface. In other words, we won't even know what he finds out from such samples. But for sure, those who do not fit this profile - which is 99.99% of all people - will be considered some kind of deviation from the pattern. Nevertheless, algorithms of this kind will be used more and more, and we just have to keep in mind that they have both positive and not very good sides.

The robot reader found out how we evaluate people

The robot has read 3.5 million books published in English between 1900 and 2008 - both fiction and non-fiction. These books contain 11 billion (!) Words, and the robot, that is, the machine algorithm, had to count the adjectives that characterized men or women. For example, finding the words “son,” “stewardess,” or “actress,” the algorithm looked at which adjectives were describing them. As a result, a list of the most frequently used adjectives, both positive and negative, was compiled, which the authors have characterized men and women over a hundred years.

The main find of this study, conducted by an international team of scientists, is that adjectives describing women mainly refer to appearance, and adjectives for men refer to behavior or personality characteristics

Most common adjectives:

Women		Men
Positive adjectives	Negative adjectives	Positive adjectives	Negative adjectives
Beautiful	Shabby	Fair	Unsuitable
Pretty	Unkempt	Healthy	Unreliable
Chaste	Barren	Virtuous	Disobedient
Gorgeous	Grumpy	Rational	Clingy
Prolific	Unsettled	Peaceful	Cruel
Lovely	Rejected	Gifted	Idle
Sexy	Unmarried	Brave	Unarmed
Chic	Undernourished	Important	Wounded
Refined	Thin	Reliable	Fanatical
Live	Uncomplaining	Sinless	Unfair
Bright	Grumpy	Noble	Ruthless

Found adjectives that describe you?

Leader of the research team, Isabelle Augenstein of the Department of Computer Science at the University of Copenhagen, notes that the study raises a number of important questions. The algorithm that was used in this search is used for speech recognition systems and will be used by other machines to communicate with people. And if there are preferences or distortions in our language, then they will certainly be transmitted to machines.

For example, a description of a woman that does not contain a single adjective related to appearance, but only characteristics of her behavior or character, the algorithm would consider deviating from the norm. Therefore, according to scientists, algorithms should be taught to ignore these biases.

But this only seems simple, because these prejudices and preferences did not appear for nothing. This should not sound like an excuse for sexism, but we, in general, all in one way or another evaluate women by their appearance, and men by their character.

Today, after thousands of studies, we know for sure that men evaluate a woman according to hundreds of parameters, almost instantly and almost always unconsciously, in the choice for procreation. This applies to everything: the characteristics of the figure, the ratio of body parts, voice, condition and color of hair, and more.

If we take only the eyes from this list - this concerns the color of the sclera, the size and color of the iris, the presence and thickness of the limbal ring, the ratio of the visible part of the sclera to the iris, and so on. These are all characteristics of youth and health. And none of us is able to consciously see, analyze and make a decision. All this is done by our internal biological algorithm, which, like it or not, has been successfully working for hundreds of thousands of years. Thanks to him, in particular, your parents chose each other and now you are reading this.

We can say, “Let's stop taking this into account. Let's turn off one algorithm for the sake of another. " This is a bold and worthy attempt, but time will tell whether we succeed or not

How do you choose the right partner?

Correct solution to the problem = stable marriage

The choice is always difficult, even if you choose one of the two. But when it comes to many participants, with incomplete information and multiple preferences, it becomes incredibly difficult. Mathematicians regularly tackle such problems, and one of them, about the choice of suitable couples, called "the problem of stable marriage" has been solved and deserves our attention.

Imagine that there are four women gathered: Anna, Bella, Vera and Galya - and four men: Alexander, Boris, Victor and Grigory. They want to create stable couples, but "everything is mixed up in the Oblonskys' house." Anna likes Boris, and he likes Galya, who likes Alexander, and that likes Vera. How to achieve stability of couples, so that your partner does not run away to another? Couple stability is when there should be no man or woman in a couple who would mutually prefer to be with others.

To figure it out, they decide to write a list of the most desirable candidates in descending order. Here's what they did.

Among women:

Preferences	Anna	Bella	Vera	Galya
1st stage	Boris	Alexander	Gregory	Alexander
2nd stage	Alexander	Gregory	Victor	Boris
3rd stage	Victor	Victor	Boris	Gregory
4th stage	Gregory	Boris	Alexander	Victor

In men:

Preferences	Alexander	Boris	Victor	Gregory
1st stage	Vera	Galya	Bella	Anna
2nd stage	Bella	Anna	Anna	Galya
3rd stage	Galya	Bella	Vera	Bella
4th stage	Anna	Vera	Galya	Vera

First round

Now, after compiling the list, the ladies begin to make proposals to the gentlemen, starting with the first on the list, and this is what the men get:

Alexander	Boris	Victor	Gregory
Bella Galya	Anna		Vera

We see that Victor is being ignored for now, while Alexander has two proposals at once: from Bella and Gali. Bella for him is in second place in preference, and he chooses her, refusing to Gala.

Second round

Galya crosses out her first candidate and proposes to the second on the list, Boris.

Boris now has two proposals, and Galya is in first place in his preferences, and he refuses Anna.

Alexander	Boris	Victor	Gregory
Bella Galya	Galya Anna		Vera

Third round

Anna, rejected by Boris, crosses him out and makes an offer to the second on her list, Alexander. Alexander again has two candidates, but Bella wins this fight and remains on the list!

Alexander	Boris	Victor	Gregory
Bella Anna Galya	Galya Anna		Vera

Fourth round

Anna proposes to the third candidate on her list, Victor:

Alexander	Boris	Victor	Gregory
Bella	Galya	Anna	Vera

Since no one rejects anyone, and there are no moves left, we get stable pairs. The more participants, the more rounds there can be, but sooner or later everyone will find the most suitable pair.

Here, we saw how one of the algorithms works. The problem looks like a parlor game, but today this algorithm works to assign children to kindergartens in Denmark, schoolchildren to schools in Hungary, students to universities in China, Germany and Spain, to assign rabbis to synagogues in New York, to connect suitable organ donors with patients in the UK. And of course, the algorithm is used by some online dating sites.

This algorithm was awarded the 2012 Nobel Prize in Economics

David Gale and Lloyd Shapley created it back in 1962. Gale died, and Shapley shared the award with Alvin Roth, who saw the potential for applying the algorithm to social problems.

You may have noticed an interesting feature: in our game, the ladies were the first to make proposals. Would something change if the offers were made by gentlemen? Surprisingly, yes. You can check it yourself, especially since it only takes one round. After him, only Boris and Galya remained together, as in the first case, and all the others created other pairs. In both cases, pairs are formed stable, and this is important.

But look what happens: when the gentlemen are the first to offer to talk, they get the best, most preferred option for themselves, and the ladies are content with slightly worse options for partners. When the ladies are the first to propose, they get better options than the gentlemen.

During the assignment to hospitals, medical students knew about the work of the algorithm and noticed that they were the first to make proposals, worsening their choice. They successfully campaigned against this approach and won the right to be the first to make offers to hospitals.

Algorithms are already working in many areas of our life, even if we are not aware of it, and will be used more and more actively. We do not need to be intimidated by this, because they are potentially created to facilitate and improve our lives. They can do something for us that is more effective than our intuitive or random choices.

Algorithm-driven online dating helps create more resilient couples than other methods. The algorithm for connecting patient and donor has already saved many lives today. And we need to know about their existence, use them and not forget about our interests

More about this:

• An article on the use of AI in hiring in The Telegraph: telegraph.co.uk/news/2019/09/27/ai-facial-recognition-used-first-time-job-interviews-uk-find/?WT.mc_id = tmg_share_tw
• Book Reading Study: copenlu.github.io/publication/2019_acl_hoyle/
• And one more thing: futurity.org/adjectives-gender-descriptions-books-2143682-2/
• The researchers used the Google Ngram Corpus database.
• The story about the "problem of stable marriage" is told in the book: Du Sautoy M. The Creativity Code: Art and Innovation in the Age of AI. London: 4th Estate, 2019.