Are decision-making algorithms always reliable?

Are decision-making algorithms always reliable?

Question

discussion
You will be writing a report or explanation, presenting the topic of the article, the nature of the argument, the ethical theory applied, and applying technical knowledge, theory, and perspectives gained in your coursework as appropriate. Criticism must be provided. Task resources. Algorithms are now more powerful and machines can use them to make decisions based on historical data and previous analysis. The analysis is mainly based on the negative results of the algorithm. Large companies and farms use various algorithms to train datasets to extract useful information that can financially strengthen their organizations. The authors’ main goal is to link decision-making algorithms with human decision-making systems. Because, according to many articles and many people’s opinions, the decision of which machine to use can fall under racist, sexist, and classist bias. Machine learning models or algorithms are now widely used to predict future events and also for predictive purposes.

There are many cases where decision-making algorithms have been found to perform worse than human decision-making and cannot guarantee justice. One such incident was witnessed at the medical department of St. George’s Hospital in London, where the output obtained from the machine was found to be sexist and xenophobic. This came as quite a surprise to them, as they had never expected a computer to be able to tell them apart. It all happened when the historical dataset used to train the model itself consisted of sexist and xenophobic elements. The discrimination predicted by the model was present according to the pattern in the dataset.

Obviously, this article is based on a comparative discussion. All algorithms are assumed to have a heart. Only a few of the algorithms are not very difficult and cannot be completely understood. Data scientists are used for this machine learning and various exploratory analysis purposes. For these data scientists, there is a need to have a clear understanding of trade-offs and build tests, as well as to monitor how the algorithm works on an ongoing basis to ensure that decisions are consistently confirmed by the algorithm. .

One of the incidents in Australia last week was when the Department of Human Services introduced an automated debt collection system that uses a crude fraud detection algorithm to calculate whether Australians are being overpaid by the welfare system. . And it turns out that fraud detection algorithms assume that all recipients’ incomes are constant throughout the year, whereas for individuals it often fluctuates from time to time. There are over thousands of complaints where authors allege flaws in the algorithm. The errors in the algorithm were not carefully looked into and also, they were not resolved due to which it increases the false positive rate for the classification model.

According to the authors point of view algorithms are not always gives what exact the outcomes would be instead using such statistical and mathematical models the author wants to state that using such tools and technology humans apparently building a weapon which will affect the humanity itself. There are many new powerful tools nowadays which were not used previously and hence in old days decisions are been made more wisely by human brains but now with such evaluation in technology many mathematicians recognized that these particular tools and model could results into some mathematical weapon.

Also building such algorithms are costly as to find the optimize algorithm different layers need to be added to an already made complex layer and need to be evaluated repeatedly if the multiple layer model or the algorithm produces optimum result or not. Moreover, and more importantly, it’s expensive because on any reasonable definition it will generally cut down on profit to be nondiscriminatory. The author is totally against of these algorithms as the author states that these decisions making algorithm cannot be trusted all the times as in real-life decision-making procedures are common but they are good unless the data fits into them and more often they are quite worse.

So, from here the question arises that if these accountable algorithms can truly do better than humans and create better and fairer decisions to provide the exact outcomes then at first data scientist have to prove using such algorithms as most of the times these algorithm fail miserably as because simply the outcomes are been generated by machines. Also, there is not neutral algorithm thus the complexity of the algorithms cannot be reduced anyhow if optimize solution is needed. Also, data collection plays a crucial role as the outcomes of any predictive algorithm depends on the data fetched into the model, if the data is well and good then the result will be accordingly and if the data contain some discrimination factor then the outcome will also have discrimination factor which can result in worse decision making ever in comparison with human decision.

Why We Need Accountable Algorithms. (2020). Retrieved March 13, 2020, from https://www.cato-unbound.org/2017/08/07/cathy-oneil/why-we-need-accountable-algorithms

“Neutral” is not neutral. (2020). Retrieved March 13, 2020, from https://www.cato-unbound.org/2017/08/11/laura-hudson/neutrality-isnt-neutral