Cognitive Technologies in Risk Management: Advancements and Challenges
Cognitive Technologies in Risk Management: Advancements and Challenges
Cognitive Technologies代写 This paper will focus on evaluating the use of AI and ML in risk management in reference to JPMorgan Chase Bank.
The use of artificial intelligence is one of the key drivers transforming the banking sector.
Today, most banks ranging in sizes in the financial sector are integrating AI not only to save on cost and boost revenues but most prominently identify risks and frauds in the banking system. (Fethi, and Pasiouras, 2010, p. 190) There are several key drivers to evolution in the sector. The intense competition coming from traditional banks and agile FinTechs and digital-only banks have necessitated most banks to utilize big data using AI and Machine Learning to attract more value for customers who have now become tech-savvy. (Giudici, 2018, p. 1)
The global economy is at the new focus on automation, big data analytics and innovations and hence finance require to adapt to this agility to fit in the digital ecosystem. The modern generation of customers is becoming more digital and spend much time online. Most importantly is the risk mitigation factors that come with the use of AI and ML and hence reduce human errors and costs. Overall, as more businesses and people use digital technologies, the business environment has become more complex and volatile than ever before. Cognitive Technologies代写**范文
However, it is a matter of what banking institutions are harnessing the power of advanced technologies. Such as AI and ML to manage and mitigate some of the risks that are prevalent in the sector.
This paper will focus on evaluating the use of AI and ML in risk management in reference to JPMorgan Chase Bank. It will also analyze challenges in the application of these technologies and how they can be mitigated for the achievement of their full potential. In so doing, the paper will start by explaining the application of AI and ML in JPMorgan Chase Bank and how it has been effective before delving on challenges and recommendations. Overall, it seeks to prove the importance of using AI and ML technologies in risk management.
Application of AI and ML in JPMorgan Chase Bank Cognitive Technologies代写
In this context, JPMorgan is a good example of modern financial institutions that are using AI and ML to manage its financial transactions (Davenport, 2019). Just like any other banking institution in America and more in the world, JPMorgan had more pressure to increase efficiency, deliver value, and at the same time improve revenues and assets. It has invested heavily in its own cognitive technology called Contract Intelligence (COiN) (Khmiadashvili, 2019, p. 86). In essence, the technology is used in various forms including chatbots, profiling customers, spot patterns in transactions, streamline the process and most of all risk management (Burgess, 2018, p. 73). Cognitive Technologies代写**范文
With more than 250 thousand employees and over $2.6 trillion assets, the integration of machine learning was essential to mitigate the most prevalent risks such as human errors and frauds. The COiN is used to analyze commercial credit agreements that mostly take 360 thousand hours of human capital complete. The use of technology reduces the serving time to few seconds and eliminates middle office processes. More so, it not only increases the bank’s efficiency but also eliminates human errors that inherent to human labor.
COiN machine learning Cognitive Technologies代写
Nonetheless, the most critical areas of application of COiN machine learning are risk management posed by frauds, cybercrimes, compliance, and lending. (Khmiadashvili, 2019, p. 87) Before Global Financial Crisis, line managers were solely responsible for risk detection and accountability but they have proved ineffective in the modern realm of digital technology and surge in transactions. JPMorgan is using its ML platform to identify risks that frauds. The platform has ability to profile clients and learn the patterns in transactions particularly in credit management. The financial crisis of the previous decade gave financial institutions many challenges in accessing the creditworthiness of their customers. Cognitive Technologies代写**范文
Before the digital revolution in financial sectors, customer profiles for creditworthiness were based on simple heuristics such as customer value data generated from focus groups and surveys consumer behaviors which did not yield much about the reality. The application of COiN technology has enabled JPMorgan to eliminate these hurdles by harnessing big data and analytics. It is able to give the company access to a really tremendous amount of data about customer behavior and needs to be based on their digital footprints and sharing of information with other financial institutions as well as government agencies.
Credit risk Cognitive Technologies代写
Additionally, the emergence of online lending technology as alternative credit access has created more credit risks to banks. The bank is using the COiN algorithm to access credit profiles of customers and by far leveraging alternative data from social media, check-ins, GPS, e-commerce, mobile data and bill payments that are important in making an informed judgment about the customer (Khmiadashvili, 2019, p. 87). Most people today have access to the internet and are therefore making many transactions online. Their activities online reveal a lot about behavior and that is essential in predicting creditworthiness. Cognitive Technologies代写**范文
For instance, the places a person visits and the types of products or services he/she buys predict the financial capability of that person. The financial status can also be predicted using photos and the kind of background images and clothes customers wear. The predictive model in ML platforms uses complex algorithms to draw conclusions about any data from the customer and give feedback to the managers for decision making in real-time. Cognitive Technologies代写**范文
Overall, the integration of AI and ML in risk management strategically aligns with the bank’s need in handling and evaluating unstructured data. A risk manager in JPMorgan uses data analytics to access losses and mitigate their occurrence based on AI and ML model predictions instead of investing much time on risk management inherent in the operational processes. The solutions offered by AI technologies have built the bank’s trust and efficiency in data delivery for use in developing customer competency and success in strategy implementation.
Possible Challenges and Recommendation in the Application of AI and ML in JPMorgan Chase Banks Cognitive Technologies代写
The application of COiN AI, however, is faced various challenges both in its implementations, usage, and dependability. There are six elements likely to impact JPMorgan bank including interpretability, bias, feature engineering, hyperparameters, production readiness, and dynamic model calibration (Kohansal, 2019). Interpretability is most conspicuous with machine learning models. The technology is sometimes referred to as “black box.” Cognitive Technologies代写**范文
It can be a challenge depending on the model architecture for risk managers to explain and understand the results generated. It is has been found that due to lack of good comprehension, managers are likely to disregard the model recommendation because they lack the knowledge to connect the rationale of the recommendations. Most managers may not trust the model which may translate to a waste of time and investment in acquiring the technology.
Therefore, banks, particularly, JPMorgan has adopted policy decisions regarding various risks. Cognitive Technologies代写
In the United States, ML models for credits are covered by fair lending laws (Doshi-Velez, 2017). As such, the model as made to produce the reason for refusal of credit. On the other hand, a recommendation by the model to advertising to a particular customer’s mobile phone may not require the reason for the recommendation since it does not pose much risk to the bank. Nonetheless, every model used by the bank needs to comply with certain policies. Cognitive Technologies代写**范文
Fortunately, the black box connotation given to the AI and ML technologies has been reduced over time to increase interpretability (Castelvecchi, 2016, p. 20). As such, JPMorgan has invested heavily in AI and ML research and development to make improvements and innovate COiN. Depending on the model type, bank can use the following approaches. First, the use of linear and monotonic models uses linear coefficients to show relationships between various variables and the recommendations.
Secondly, the application of non-linear and monotonic model requires limiting the input variables in relation to dependent variable to simplify the model prediction. Thirdly, non-linear and non-monotonic which has unrestricted learning uses local interpretable models to increase interpretability. When these approaches are used in various combinations and in regard to the model type, the interpretability of the recommendations is improved.
At the same time, the COiN model is prevalent in bias. Cognitive Technologies代写
There are four main bias observables in various types of AI and ML technologies including “sample, measurement, and algorithm bias, and bias against groups or classes of people.” The latter is amplified in the model. The implementation of the random-forest algorithm was found to be biased toward input variables that are more distinct and hence resulting in poor decisions. (Caliskan, Bryson, and Narayanan, 2017, p. 184; Strobl, Boulesteix, Zeileis, and Hothorn, 2007, p. 25) Cognitive Technologies代写**范文
Banks need to be aware of these faults. For instance, a model developed using a random-forest algorithm to predict money laundering activities in the bank was found biased against fields with few numbers. Such as country and used them to predict tendencies of money laundering and exclude those with large numbers like occupations. The solution to this challenge is to develop a model validation system in the model and making regular updates of algorithms or create an alternative model that will be used to benchmark the model performance.
On the other hand, the model needs to be impartial with regard to groups of people and classes such as gender bias. (Leavy, 2018, p. 14) Cognitive Technologies代写
It is the role of banks to decide what constitutes fairness in risk prediction and management. Depending on the model the bank can use four definitions to manage fairness including demographic blindness, demographic parity, equal opportunity, and equal odds. The use of demographic blindness and parity means reducing the use of features that correlated to the marginalized group and results are equal for all groups respectively. The bank needs to validate that developers have used algorithms that will ensure fairness in prediction.
Moreover, banks are challenged by the lack of skills and quality data. Cognitive Technologies代写
The financial industry is plagued with the problem of reconciling the data from front to back and that data from other sources have quality issues. (Thrall et al. 2018, p. 505) Though JPMorgan bank has invested heavily in AI and ML, it had to also invest in human resource development toward the adoption of the technology. The reason being there are limited numbers of staff that well acquainted with the effective use of AI and ML. Lack of skilled labor is topped up by a lack of adequate and usable data. The integration of these technologies become expensive and time-consuming. Cognitive Technologies代写**范文
Therefore, just like JPMorgan bank other financial institutions are investing heavily in research and development to produce niche applications where data is readily available. There is also focus on data harvesting to have adequate amount of data for decision-making as well as training and development of employees to adapt and be able to use and apply these technologies. The ability to acquire adequate quality data and having skilled human labor is essential in implementing large-scale artificial-intelligence initiatives.
Furthermore, cognitive intelligence technologies are designed to solve specific problems and are not capable of solving anything outside its learning. (Thrall et al. 2018, p. 505) Cognitive Technologies代写
For instance, COiN has an algorithm designed for credit contract profiling so that the company can reduce the chances of errors, risks, and frauds in lending contracts. Though it can have other modules, it cannot be used for online money transfer. Also, an algorithm created to detect suspicious payments cannot be used to detect other suspicious activities in transactions. Additionally, AI and ML are designed to be rational in decision-making and hence lack emotional intelligence and contextualization of information like human beings. Cognitive Technologies代写**范文
The chatbot in COiN ML sometimes can disappoint because it lacks empathy in conversations. As such, banks should be aware of these shortfalls and to not fully leave some decisions that require emotional judgment to AI/ML. The decisions from these technologies should ultimately be under human discretion.
Lastly, the application of cognitive technologies presents a challenge of liability when something goes wrong (Thrall et al. 2018, p. 506). Cognitive Technologies代写
As a result, financial institutions have not given AI/ML full autonomy in behavior and decision-making. The reason being, their behaviors are not fully predictable. A real example of unpredictability and failure of AI was witnessed by IBM’s “Watson for Oncology” when it was canceled for giving unsafe treatment recommendations (Herper, 2017). Another notable failure in AI and ML is Microsoft’s own chatbot Tay which was corrupted by the internet troll within 24 hours. It was flooded with a deluge of racist, misogynistic and anti-Semitic tweets. That it was turned into a mouthpiece for a terrifying ideology (Vincent, 2016). Cognitive Technologies代写**范文
These and other examples in various industries have kept banks from fully rolling AI and ML as independent decision-makers. Therefore, human supervisors are used to validating machine decisions for critical activities such as making payments and verification of payments. These shortcomings in AI and ML usage have resulted in high regulations and compliance standards as well as operational security measures in the anticipation machine errors until such a time. That the technology will be mature enough to be autonomous.
Despite few challenges that are inherent to the use and application of AI and ML Cognitive Technologies代写
In broader look of its contributions not only to the banking sectors but also to all aspects of life. It is a revolutionary technology that is long overdue. Banks have leaped a lot from its integration in various functions yet the technology is at development stage. They able to handle bulk of transactions and solutions at once which was not possible with human limitations. Costs resulting from errors, frauds and other internal controls are significantly reduced to the minimum and manageable levels. The skepticism that exists will tremendously end as the technology advance to solve some issues identifiable with its use.
References Cognitive Technologies代写
Burgess, A., 2018. AI in Action. The Executive Guide to Artificial Intelligence (pp. 73-89). Palgrave Macmillan, Cham.
Castelvecchi, D., 2016. Can we open the black box of AI?. Nature News, 538(7623), p.20.
Caliskan, A., Bryson, J.J., and Narayanan, A., 2017. Semantics derived automatically from language corpora contain human-like biases. Science, 356(6334), pp.183-186.
Davenport, T., 2019. AI At JPMorgan Chase—Breadth, Depth, And Change. Forbes. Available from https://www.forbes.com/sites/tomdavenport/2019/11/12/ai-at-jpmorgan-chasebreadth-depth-and-change/#43aebd6a110b
Doshi-Velez, F., Kortz, M., Budish, R., Bavitz, C., Gershman, S., O’Brien, D., Schieber, S., Waldo, J., Weinberger, D. and Wood, A., 2017. Accountability of AI under the law: The role of explanation. arXiv preprint arXiv:1711.01134.
Fethi, M.D. and Pasiouras, F., 2010. Assessing bank efficiency and performance with operational research and artificial intelligence techniques: A survey. European journal of operational research, 204(2), pp.189-198.
Giudici, P., 2018. Fintech Risk Management: A Research Challenge for Artificial Intelligence in Finance. Frontiers in Artificial Intelligence, 1, p.1.
Herper, M., 2017. MD Anderson Benches IBM Watson In Setback For Artificial Intelligence In Medicine. Forbes. Available from https://www.forbes.com/sites/matthewherper/2017/02/19/md-anderson-benches-ibm-watson-in-setback-for-artificial-intelligence-in-medicine/#5a90705f3774
Kohansal, M., 2019. The Evolution of Model Risk Management: for Managers and World Leaders.
Khmiadashvili, L., 2019. Building progressive future: human–AI collaboration. Calitatea, 20(S3), pp.85-88.
Leavy, S., 2018, May. Gender bias in artificial intelligence: The need for diversity and gender theory in machine learning. In Proceedings of the 1st International Workshop on Gender Equality in Software Engineering (pp. 14-16). ACM.
Strobl, C., Boulesteix, A.L., Zeileis, A. and Hothorn, T., 2007. Bias in random forest variable importance measures: Illustrations, sources, and a solution. BMC bioinformatics, 8(1), p.25.
Thrall, J.H., Li, X., Li, Q., Cruz, C., Do, S., Dreyer, K. and Brink, J., 2018. Artificial intelligence and machine learning in radiology: opportunities, challenges, pitfalls, and criteria for success. Journal of the American College of Radiology, 15(3), pp.504-508.
Vincent, J., 2016. Twitter taught Microsoft’s AI chatbot to be a racist asshole in less than a day. The Verge. Available from https://www.theverge.com/2016/3/24/11297050/tay-microsoft-chatbot-racist