top of page

Discussion Topics.
Download underlying documents in the knowledge bank

Behavioral ethics and Artificial Intelligence

Behavioural Ethics is based on the notion of bounded ethicality (see below), which in turn refers to being limited in ethical awareness, This resembles the notion of bounded rationality by H. Simon. Wikipedia: "With bounded rationality, Simon's goal was "to replace the global rationality of economic man with a kind of rational behavior that is compatible with the access to information and the computational capacities that are actually possessed by organisms, including man, in the kinds of environments in which such organisms exist."

 

Bounded rationality has led to behavioural finance, a rapidly growing and successful field of research, acknowledging the limitations in our attention. Similarly, we could speak of behavioral ethics to refer to a rationality that takes into account the limitations in availability of information and capacity for processing it by studying actual behavior as the product of such limitations.

 

An example of a source for unethical behavior is the bias that can result from seeking confirmation of decisions made, known as confirmation bias. This necessitates a-priori statements of goals to be achieved in such detail that progress can be measured.

 

The increasing use of Artificial Intelligence, and particularly Machine Learning (AI/ML), has raised attention for the ethical aspects of modelling and applying the results to people, when this is performed by machines. Computer says no.

First experiences in ethical assessments of algorithms, or models, indicates that intentions are often mentioned as arguments for being ethical. Our ambitions are good, so we do good. Although often the starting point, good intentions are not sufficient by themselves. As the proverb states: the road to hell is paved with good intentions.

A model needs to perform as intended, and evidence (monitoring instead of ambitions) of that is required for assurance. Behavioral ethics urges to focus on actual behavior/performance of people and algorithms, not on ideals or ambitions.

Bounded ethicality

From scientific research[1], it is known that people (e.g. employees) apply a “bounded ethicality”. The first finding of this research was that individuals are more likely to lie, or commit fraud, when they are set excessively difficult and specific goals. Too ambitious goal setting is a source of bias. Bounded ethicality, the authors argue, can also operate at an unconscious level. Under pressure, people often do not efficiently analyse information that could otherwise keep them on the straight and narrow.

The problem is exacerbated by confirmation bias, a human tendency to seek out facts that back up their pre-existing preferences. Research has found that people given a specific performance target (to reach a 12% annual return over the investment horizon, for example) were more likely to overlook important information about the future performance of investment funds and excessively focus on past performance data.

As a result, the authors suggest, “Organisations might decrease intentional unethical behaviour by defining their goals more broadly and by setting goals at levels that are perceived as fair and relatively attainable by employees.” Another tactic is for managers to signal clearly that ethical issues may arise, so that people take them into account when making decisions. Commitment to a decision made them less likely to attend to ethically relevant information, i.e. after their decision had been made, individuals were more prone to ignoring information about the unethical behavior. This means that ethical awareness has to be raised before committing to a decision, i.e. a priori.

What can we do to mitigate the risks to acceptable levels?

Below, a number of actions are detailed:

  1. Specify upfront any ethical dilemmas that may arise and compare them with moral norms and values of the business in the assessment of required knowledge.

  2. Propose counteracting actions to mitigate the risks associated with each of the ethical dilemmas where required, including ex-post monitoring of model performance and emergent bias.

  3. Organise an explicit approval of the ethical assessment and risk acceptance of the expected biases.

 

[1] Reducing bounded ethicality: How to help individuals notice and avoid unethical behaviour,” by Ting Zhang, Pinar Fletcher, Francesca Gino and Max Bazerman, https://dash.harvard.edu/bitstream/handle/1/27419676/zhang%2Cfletcher%2Cgino%2Cbazerman_reducing%2Bbounded%2Bethicality.pdf?sequence=1&isAllowed=y

behavorial ethics

Philosophy of financial risk models

Increasingly, financial risk models are applied in the business environment. Especially banks use risk models to perform main business functions, such as risk management, client selection and deal structuring, capitalisation, performance measurement, etc. 

 

The current mainstream financial risk models are quantitative, based on statistics, and rely heavily, as a condition sine qua non,on computing power and communication networks., such as inter- or intranet. This technology is relatively new, and still developing rapidly (Moore’s law says a doubling of computing power every two years).

Artificial Intelligence, robotics, deep learning is flowing into bank’s offices, pushed by technological developments. The technological asset is developing fast.  However, the human factor, currently still the initiator of technological drive, is much slower in its developments. Understanding of the new technology is slower than the technological development itself.

Mainstream models are based on statistical theories that were developed in the 17thcentury[1]or based on rational actor theories that are Modernist (based on Descartes, 1648). In recent years, the most prevalent version of rational choice theory, expected utility theory, has been challenged by the experimental results of behavioral economics. Economists are learning from other fields, such as psychology, and are enriching their theories of choice in order to get a more accurate view of human decision-making. For example, the behavioral economist and experimental psychologist Daniel Kahneman won the Nobel Memorial Prize in Economic Sciences in 2002 for his work in this field. However, proliferation of these new ideas is slow.

In this article, an attempt is made to philosophy about the development and application of financial risk models in order to gain more understanding about the implications of using current mainstream models. This is done in the aftermath of the largest financial crisis in the life of the author, which, not coincidentally happened in the year that financial risk models were used first in banks to calculate required (or regulatory) capital (known as Basel II).

Lay out of the article

  1. Risk management: 1) awareness, 2) counteraction, 3) evaluation (measurement). 
    In this article we work on 1) awareness.

  2. Description of current main stream credit risk models ito epistemology & methodology Empirical statistics

  3. Nozick’s description of what this (“probability example”) means 

  4. Negligence of expertise whilst it is fully used in all elements of the chain

  5. Underdetermination by data

  6. Data quality

  7. Are we happy with our models? Positivity metaphor. Reflectivecharacter. 
    What can we do ourselves?

 

[1]In 1654 Pascal and Fermat discovered the theory of probability, the mathematical heart of the concept of risk.

ph of financ ris models

Philosophy of travelling

The philosophy of travelling is based on an alternative conceptual morphology of exchanges, dynamics and flows, instead of objects, stability and facts. The philosophy of travelling provides a framework for historical analysis of society (and its main type of travelling), of epistemological and methodological notions, as well as a framework for current social analysis. In my view, this perspective explicitly deals with dynamics and uncertainties and is therefore most suitable to analyse risk theories.

In the genealogy of travelling, 4 types of travelling can be distuinguished:

  1. the hunter, the first type, travelled where his food went, not being able to grow his own food. The hunter only has what he can carry with his nomadic lifestyle.

  2. the farmer, who learned to grow and harvest his own food, which requires him to stay on his land, but also enable to create abundance and save this in warehouses. The farmer in fact does not travel through lands, but stays, builds fences and defends his place and creates order.

  3. the merchant, who travels between supply and demand, using money to connect to different environments he crosses, buying, producing and selling.

  4. the informatician, who travels only digitally via ICT and internet, and is actually everywhere, accommodated by the ultra speed of the information processing. He does not rely on gold, but on information, requiring a new approach.

Phil of trav

Pragmatic Information Theory

The Pragmatic Information Theory is a newly designed concept intended to help managers to improve the value of their company’s information, the most important asset.

It is pragmatic because it deals with actual issues in day-to-day management life and focused on use of information in applications as criterion for success.

It is a new view on information based on the virtual character of it, as opposed to the real (physical world) character of things like money or matter. Information is referential, reflexive and is multiplied when divided, hence needs a different approach.

As the focus is on use, the concept of Chain of Information is presented in this article, which can be used as a tool to actually manage the value of specific information flows. It helps to identify the weak spots in the chain. Improvement of the weakest spots increases the value of the chain.

Next, a first application of the chain-concept in portfolio management within a bank is elaborated as illustration of its practical use.

Finally, adjacent theories that have been inspiring, such as cybernetics, risk society and asymmetric information, and their relation to Pragmatic Information Theory are described.

PIT

PhD Thesis:  

The EC movement, or the quantitative turn in banking, a case study of the virtual turn of society

In this thesis we have investigated a (Rabobank) specific case of the EC movement, which is in itself an example of the virtual turn in society. After spending time to develop an approach, the thesis presents three parts of increasing generality, revolving about three research questions:

 

  1. What are the characteristics and weaknesses of EC theory?

  2. What are the social implications of the EC movement?

  3. What are potential implications of the virtual turn in society at large?

 

The idea behind this set of questions is that society might be turning virtual; a transformation of society from monetary to informationalist society. The quantitative turn in banking, a.k.a. the EC movement, is a good example of this virtual turn in a specific segment of society. However, the banking sector is a core segment in society, a spin in its web. If we study the EC movement, we find out how this virtual turn takes effect in banking, and we might be able to extrapolate findings to a wider context than just banking.

In order to study the EC movement, a clinical case study was performed, ie. the Rabobank implementation of the BIS II/EC framework was used as a case of the quantitative turn in banking.

 

Investigation of literature in philosophy of science showed a postmodernist crisis of science and philosophy of science. Recently it has been shown that objective knowledge is not an adequate picture for scientific development, since testing is partial, goals for proper science conflict, context plays a decisive role, human interaction is important, etc. A set of complicating factors has been identified which complicate the formal (modernist) notion of science. Standard recent philosophy of science suggests a classification scheme for analysis of theories, attention for contextual factors as well as a multitude of perspectives. Nozick, who provides a functional response to post modernism, holds that rationality of knowledge is in the counteraction of biasing factors, such as the complications mentioned above. Finally, the philosophy of travelling was presented, based on an alternative conceptual morphology of exchanges, dynamics and flows, instead of objects, stability and facts. The philosophy of travelling provides a framework for historical analysis of society (and its main type of travelling), of epistemological and methodological notions, as well as a framework for current social analysis. In my view, this perspective explicitly deals with dynamics and uncertainties and is therefore most suitable to analyse risk theories.

 

Part one consists of a classification of EC theory and an investigation into specific issues in the content of the theory. The classification of EC theory showed that the EC movement develops in a regulated business context, and is driven by (ICT) technological developments, risk theory sophistication and financial markets developments (improving liquidity in credit financial instruments). Direct stakeholders are the Regulators, the shareholders, and credit risk management, who all expect significant improvements from EC. Furthermore, an ontological analysis of capital and credit risk showed that EC is a blurred concept about rare events. An epistemological analysis showed that formal or mainstream EC movement was based on empiricism, statistics and econometrics, while the Global Financial Markets provided information about reality for more advanced practitioners. A remarkable inconsistency was noted between the formal epistemology and the ontological nature of the subject of the theory. The subject is blurred and rare, while the formal school assumes the Law of Large Numbers. Next to that, risk is essentially normative and affective, therefore subjective, while the formal school declines any subjectivity and assumes objective risks.

 

Turning to issues in the content of EC theory, observations were made regarding risk concepts, diversification of risks, valuation principles and quantification.

The study of the concept risk showed that risk is an old concept, but calculated risk, with its cornerstone of probability, has emerged as a concept after the middle ages, when man actively started to rule his own life, breaking away from religion or tradition. Risk requires causal structures to generate expectations, and normative statements to identify risks and acceptance levels; is subjective in terms of related to specific expectations, evaluations and performances, but also objective in terms of outside one person, accessible to third parties as well. Conceptually, portfolio credit risk –at the centre of EC theory- is not a well understood concept, borrowing concepts from capital markets, and assuming perfect markets. Next to that, we see that the prerequisites to apply the concept of calculated risk are failing in contexts of wholesale finance. Finally, in the company of regulators, we find that validation of the models is a weak point, with the possibility that there may never be adequate amounts of data.

The concept of diversification essentially says that a portfolio should not be concentrated, should be as diversified as possible. Adding volatility with less than 100% correlation with the portfolio reduces volatility, reduces risk. However, empirical studies suggest that diversification into unknown industries is a bad idea. Next to that, the Rabobank’s credit policies require significant expertise concerning viability of the client, its industry, etc. This seriously restricts the available space to diversify. Finally, two examples of industry diversification from the bank’s recent history were mentioned as cases where diversification did not yield promised results. 

From a valuation perspective, capital is quite a blurred concept. It comes from different sources, is valued as a Net Asset Value, and thus subject to all valuation issues for all assets and liabilities. A significant issue concerns the accounting mismatch between banking and trading books, which makes that loans are valued by their cost price, while the hedges of the loans are valued by current market price.

The quantification issue revolves about the need for and constraints of quantification. EC theory quantifies risks by applying a central standard and numerical language, formatted on the econometric parameters of average, volatility and loss distribution. We find that the quantification is required in order to apply the benefits of mathematics and computers. Numbers can attain the speed of light in ICT systems. The link with strictly empiricist / statistical methodology creates the bottle neck of this approach: the indefinite need for objective data. Any proper level of differentiation in the model will be impossible to validate statistically. Expert based quantification provides an alternative.

 

As a final conclusion of this part of the thesis, it is noted that the weaknesses in the content of EC theory seem mostly attributable to the inconsistency between ontological nature of the subject and choice of epistemology/methodology. Risk is subjective, not an objective probability, it is not empirically measurable.

 

In part two, the social implications of the EC movement have been investigated. The EC movement is based on three fundamental developments that change society in a wider context than the EC movement. However, the EC movement also provides a strong impetus to these fundamental developments; it makes respective investments worthwhile. These fundamental developments include:

  • ICT developments, creating a network society with speed of light technologies. The network society changes relations of productions, of power and of experiences.

  • Risk management sophistication, creating a risk society in which awareness for the side effects of production becomes larger than for the production itself. Risk society is reflexive in terms of producing its own risks, fighting itself, being emancipated from natural hazards.

  • Global economy, with its nerve centre: the global financial markets. Capital mobility has increased considerably, and the EC movement stimulates further acceleration. However, the global economy has many negative side effects. Since the EC movement is (also) based on privately owned technology, it will in first principle not improve global capitalism, and likely stimulate it further.

 

The philosophy of travelling was applied to identify the major types of travelling as ICT, Risk Management, Global Capitalism, and EC movement specific travelling. ICT pulls us into the space of flows, where distance is determined by connection to flows, and timeless times, where all times are available simultaneously in the network. Risk management focuses us on the future, shows us how to appropriate the future. When risk management concerns issues that affect society, risk management becomes a political vehicle. Global Capitalism provides hyper mobility to equity, creating speed differentials with other business resources. Next to that, it provides an exit for capital, so it might run away for its responsibilities. EC movement travelling favours concentration of the industry, which is somehow contradictory with the adagio to diversify.

 

In part three, first the concept of virtuality was investigated in order to clarify what is meant with a virtual turn in society. Next, findings from the clinical case and the investigation of social implications were summarised and generalised in terms of its two main issues.

 

Origins of virtuality date back to Plato, with his theory of Ideas, or even before that, to the first use of the Word. As is shown, new in this virtuality is not the spoken or written word, but the digital word. Only the digital world can reach the ultra speed of our ICT. 

Virtual travelling is technologically supported, hence the interactions are mediated by technology, mixing environments and de-linking time and place, with superior processing speed and external storage capacity and putting human judgement on a distance, since the machines need algorithmic rationality.

Indications of the existence of the virtual turn were derived from business, politics, culture, private life, etc.

 

Summarising parts one and two, it was concluded that the EC movement is based on a powerful tool, based on speed of light technology and geared towards the financial markets, but sacrificing many human aspects. For example, in the recognition that risk assessment is essentially subjective, that the bank has its own role in the developments of risks, etc.

Generalising the conclusions, we find that two issues must have a major impact on the orientation of (people in) society; the ultra speed of virtual travelling, and related to that, the breaking of traditional horizons based on locality. Questions pop up regarding how to give meaning and feel solidarity with other people; how to produce if we all turn into nomadic warriors; should we constrain technological possibilities, or constrain where our attention goes to, and if so, how?

 

In my opinion, one should apply a proper epistemology, adequate for the nature of the subject. The approach should fit the circumstances. This means that human judgement must be included in the formal approach, or there will be an informal approach that guides our actions, and a formal approach for the regulators or other formalists. Next to that, I believe that the involvement of people also produces commitment and democratic participation in the design of tomorrow’s world. Given the constraints of current technology and concepts, there is a lot to win.

 

At the end of this thesis, it is clear that the virtual turn –and in particular its design and fundamental principles, will change our lives. When turning virtual, we should ask ourselves: does the technology do all the thinking, or should we keep the essential judgements to ourselves? After all, only humans can give meaning to life.

PhD thesis summary

© 2021 Adriaan Kukler

bottom of page