Kozminski Techblog

Nowadays, Artificial Intelligence is developing at an unbelievable pace. It has become literally omnipresent being incorporated into numerous industries all over the world. But there is one significant drawback that prevents AI from reaching its perfection – the presence of bias.  So, what is the reason for this defect in such innovative technology? Since AI is based on millions of data sets created by humans, it digests an enormous and diverse amount of people’s preconceptions, prejudice and, of course, biases. So, in what way is biased AI manifested? To provide an example, I want to elaborate on some of the cases, that constituted a red flag for AI developers. The example that comes to my mind first is the issue that face recognition systems were undergoing while misidentifying the faces of people of color more frequently than the faces of white people. Another great example depicting the margin in the AI system is the prevalence of male pictures in Google images that appear when a user enters “Image of a CEO” into the search engine. One more biased AI algorithm was spotted in Amazon’s recruitment service when the majority of women who were applying for technical or software development positions were rejected and priority was given to male applicants. There are millions of similar situations that we could recall, hence, the question arises, what is the reason for bias emergence in certain AI systems and algorithms? Is it related to technological bugs? It is not. The result being produced by AI is fully dependent on the data that is being inputted into it and later processed by it. Therefore, the single reason for bias in AI is the imperfection of the data which it relies on. 

So what measures should we take in order to mitigate the presence of bias in AI? The first step that should be made in order to tackle this issue is to improve the datasets that constitute the basis of AI systems. The lack of data is a widespread problem that data miners face and that inevitably leads to the inclinations of AI toward particular decisions. Hence, it is crucial to work only with complete databases, that include members of all gender, racial and age classes and are clear from over or underrepresented groups of people. But how to achieve this versatility and diversity of the data set? The involvement of OSDS in the projects is the answer to this question.  OSDS, which is Open Source Data Science is a database that is free to view and modify – something similar to codes being visible to all the users of the software or website. The incorporation of such a method into an AI project is a perfect possibility both for AI workers to diversify the database since all people will have access to enhance it and also for fresh data analysts to improve their skills by being able to detect bugs and improve the visible-to-all AI algorithms. Since people will be provided with open-source data sets, they will be strongly motivated to upgrade them by finding mistakes, removing biased data, adding data for minority groups or even initiating crowdsourcing campaigns when finding out that there is a shortage of certain data. In my opinion, it is an amazing collaborative approach for not only progressing the technology industry but also turning our society into a cohesive working organism being opened to innovations and developments. 

Considering the potential implication of OSDS, is it possible for AI to ever become unbiased? I am vague about it. The problem is that people who are the designers of AI systems and the preprocessors of data are naturally biased and therefore give birth to this bias in AI systems. It is impossible to deprive people of their prejudice and preconceptions and that is the reason why AI is being filled with biased data. Even bugs are being fixed by people with biased mindsets, hence it is almost impossible to impede bias manifestation in algorithms. Nevertheless, the more diverse dataset, processed by AI, would be and the more people would take part in its composition, the higher chance would be for the AI algorithm to be less biased. That is why I am strongly convinced that the application of OSDS is an approach that could elevate the development of AI to the next level by reducing the existence of biased data there. 

Sources:

https://www.weforum.org/agenda/2022/10/open-source-data-science-bias-more-ethical-ai-technology/

https://levity.ai/blog/ai-bias-how-to-avoid

https://www.forbes.com/sites/bernardmarr/2022/09/30/the-problem-with-biased-ais-and-how-to-make-ai-better/?sh=596670e14770

https://www.geeksforgeeks.org/5-algorithms-that-demonstrate-artificial-intelligence-bias/

Artificial Intelligence’s development brought numerous innovative changes to our world and still elevates the development of technology to the next level. Even though, disregarding the opportunities, evolution and new visions, that originated as a result of AI’s introduction, a considerable part of our society is enormously scared of one drawback that AI could produce – a massive takeover of jobs, which could be performed by technology instead of humans. However, being misled by the huge inflow of Hollywood movies and personal preconceptions, people do not consider that although AI will eliminate most jobs performed by humans, it will also create new professions which will be more important for our well-being. Let’s take the emergence of the internet: it was an extremely innovative invention that solely beneficially affected our population by giving birth to numerous jobs and significantly changing the existing perceptions of how people should work. The same thing is with Artificial Intelligence, its development won’t take jobs away, on the contrary, it will generate new positions and what is more – influence the existing ones. It will significantly digitalize, automate and alleviate the working processes. So, the major task of our society is to go with technology’s evolution hand in hand. 

People hypothesize that the majority of occupations will become “extinct” since they will be substituted with automation provided by AI. Of course, it is inevitable that AI and robots will exterminate some of the jobs such as cashiers, bookkeepers, couriers, accountants, etc., however, the level of unemployed people won’t increase if the population manages this transition properly. We should not concentrate on the preservation of jobs that AI is likely to take over, but to prepare our workforce for the shift to the new types of work that will be brought about by technological development. The main goal is to transmit the values of such a radical transformation to humanity. First of all, people should be aware of the economic opportunities, that would be growing in correlation with automation’s improvement. To go into more detail, automation could considerably enhance the production processes in manufacturing factories making it more rapid, cost-effective, and higher yield. The development of manufacturing and other spheres sufficiently contributes to GDP growth thus stimulating the economic prosperity of a country. In order to reach such results, more workers should be motivated to gain technical skills. Secondly, the adoption of people to the technologically-innovated world will lead to the availability of more career opportunities for them. For instance, a study, provided by a World Economic Forum predicts: “By 2025, digital technologies will create at least 12 million more jobs than they eliminate, as people with the right skills will be needed to program, maintain and repair them”. If our society is prepared for technological changes, the implementation of robotics and AI within a company will not lead to mass dismissals but will result in the transfer of employees to more crucial positions. Undoubtedly, the positions connected with technological skills and knowledge require a lot of work, devotion and consistency. But isn’t it exactly that approach to the work, which humanity should seek? 

To conclude, I am convinced that people waste a lot of time on something that could be accomplished by AI, instead of investing their efforts and dedication into the more vital activity that is likely to initiate further development of our world. They should switch from repetitive tasks, which don’t require a lot of creativity and enthusiasm, to more beneficial, rewarding and meaningful ones. Consequently, the way people perceive the essence of work will be completely changed since they will realize that technology could be in charge of automation while society should be responsible for innovation. If businesses and governments unite their efforts and make all their best to invest in the training of technologically-skilled workers,  the shift to innovative and technologically-advanced working processes will be put forward thus giving birth to a new, digital environment in which there will be a developed strategy of humans’ and machines’ collaboration, not competition. 

Sources:

https://www.weforum.org/agenda/2022/09/how-automation-job-creation-hand-in-hand/

https://www.forbes.com/sites/ashleystahl/2021/03/10/how-ai-will-impact-the-future-of-work-and-life/?sh=77c2817879a3

https://www.beyond.agency/blog/will-ai-take-my-job

The number of spheres where Artificial Intelligence successfully substitutes humans’ actions and work is considerably growing and is projected to grow at the same pace in the foreseeable future. I am genuinely impressed by the diversity of these fields. But have you ever thought whether AI could substitute judges in courts and that instead of a person you will see a robot in a courtroom? 

To begin with, the application of Artificial Intelligence in the judicial systems is already sufficiently developed in some countries. Automatic information and evidence extraction, prediction of crime occurrence and optimisation in data acquisition are one of the most common AI techniques integrated into judicial policies. However, whether AI application in courts could evolve from assistance into the replacement of humanoid judges is still an open question.

Let’s figure out how the idea of AI-driven adjudication has arisen. There are 2 main issues that have led to the consideration of humanoid judges’ inefficiency. The presence of human bias and discrimination is the first obstacle on the road to the judicial system’s perfection. It is a common situation that personal preconceptions of judges regarding the gender, race or age of the defendant could incline them to a negative court resolution. Another common problem that is widespread among judges is the correlation of personal mood and mental condition with their decision. Very often the final argument of the jury is considerably influenced by his feelings or emotions, which could be the reason for the unfair imprisonment of an innocent person or ,on the contrary, the relief of a dangerous maniac. Of course, a robotic system is not capable of being affected by a bad mood because of a broken nail or a dramatic break-up and neither is it programmed to be dependent on certain prejudices. Even though, now there is an ongoing debate among society’s members whether a robot could take the role of the judge. Although AI systems guarantee greater fairness and the elimination of bias and discrimination in the courts, numerous issues still have to be considered and fixed. The lack of emotional intelligence and accountability and moreover the inability to make unprecedented decisions set significant barriers to the transition of the judicial system to a fully technological one. Many professional data analysts are convinced, that since AI is trained by being programmed to synthesise and analyse the immense inflow of previous cases, it unintentionally processes the bias and prejudice of numerous lawyers who dealt with these cases in the past. Hence, there is no real possibility to get rid of biased decisions in courts since AI is constantly digesting bias and discrimination patterns of humanoid judges.

So, in my opinion, the application of AI in the judicial environment should be emphasised on help and assistance rather than on the total substitution of human beings at the head of the court. Artificial Intelligence could automate and alleviate the work of judges to a large extent by saving them a lot of time and providing them with a greater possibility to concentrate more on the final decision rather than on case details and documentary work. Of course, we could expect the introduction of AI to the judge’s duties, as the presence of bias, discrimination and prejudice are at least minimised there,  but before this implementation, bias and discrimination tendencies should be removed from people’s mindsets thus bringing AI judicial systems to perfection. 

References:

https://www.ft.com/content/a5709548-03bd-4f65-b9b5-7aa0325c0f6b

https://www.worldgovernmentsummit.org/observer/articles/2017/detail/could-an-ai-ever-replace-a-judge-in-court

https://contractpodai.com/news/ai-replace-lawyers-judges/