Bias and Critical Thinking

From Sustainability Methods

Note: The German version of this entry can be found here: Bias and Critical Thinking (German)

Note: This entry revolves more generally around Bias in science. For more thoughts on Bias and its relation to statistics, please refer to the entry on Bias in statistics.

In short: This entry discusses why science is never objective, and what we can really know.

What is bias?

"The very concept of objective truth is fading out of the world." - George Orwell

A bias is “the action of supporting or opposing a particular person or thing in an unfair way, because of allowing personal opinions to influence your judgment” (Cambridge Dictionary). In other words, bias clouds our judgment and often action in the sense that we act wrongly. We are all biased, because we are individuals with individual experiences, and are unconnected from other individuals and/or groups, or at least think we are unconnected.

Recognising bias in research is highly relevant, because bias exposes the myth of objectivity of research and enables a better recognition and reflection of our flaws and errors. In addition, one could add that understanding bias in science is relevant beyond the empirical, since bias can also highlight flaws in our perceptions and actions as humans. To this end, acknowledging bias is understanding the limitations of oneself. Prominent examples are gender bias and racial bias, which are often rooted in our societies, and can be deeply buried in our subconscious. I think it is our responsibility to learn about the diverse biases we have, yet it is beyond this text to explore the subjective human bias we need to overcome. Just so much about the ethics of bias: many would argue that overcoming our biases requires the ability to learn and question our privileges. Within research we need to recognise that science has been severely and continuously biased against ethnic minorities, women, and many other groups. Institutional and systemic bias are part of the current reality of the system, and I believe that we need to do our utmost to change this - there is a need for debiasing science, and our own actions. While it should not go unnoticed that institutions and systems did already change, injustices and inequalities still exist. Most research is conducted in the global north, posing a neo-colonialistic problem that we are far from solving. Much of academia is still far away from having a diverse understanding of people, and systemic and institutional discrimination are parts of our daily reality. We are on the path of a very long journey, and there is much to be done concerning bias in constructed institutions.

All this being said, I want to shift our attention now to bias in empirical research. I offer three different perspective in order to enable a more reflexive understanding of bias. The first is the understanding how different forms of biases relate to design criteria of scientific methods. The second is the question which stage in the application of methods - data gathering, data analysis, and interpretation of results - is affected by which bias, and how. Finally, the third approach is to look at the three principal theories of Western philosophy, namely reason, social contract and utilitarianism - to try and dismantle which of the three can be related to which bias. Many methods are influenced by bias, and recognising which bias affects which design criteria, research stage and principal philosophical theory in the application of a method can help to make empirical research more reflexive.

An overview on biases. Source:

Design criteria

While qualitative research is often considered prone to many biases, it is also often more reflexive in recognising its limitations. Many qualitative methods are defined by a strong subjective component - i.e. of the researcher - and a clear documentation can thus help to make an existing bias more transparent. Many quantitative approaches have a reflexive canon that focuses on specific biases relevant for a specific approach, such as sampling bias or reporting bias. These are often less considered than in qualitative methods, since quantitative methods are still – falsely - considered to be more objective. This is not true. While one could argue that the goal of reproducibility may lead to a better taming of a bias, this is not necessarily so, as the crisis in psychology clearly shows. Both quantitative and qualitative methods are potentially strongly affected by several cognitive biases, as well as by bias in academia in general, which includes for instance funding bias or the preference of open access articles. While all this is not surprising, it is still all the much harder to solve.

Another general differentiation can be made between inductive and deductive approaches. Many deductive approaches are affected by bias that is associated to sampling. Inductive approaches are more associated to bias during interpretation. Deductive approaches often build around designed experiments, while the strongpoint of inductive approaches is being less bound by methodological designs, which can also make bias more hidden and thus harder to detect. However, this is why qualitative approaches often have an emphasis on a concise documentation.

The connection between spatial scales and bias is rather straightforward, since the individual focus is related to cognitive bias, while system scales are more associated to prejudices, bias in academia and statistical bias. While the impact of temporal bias is less explored, forecast bias is a prominent example when it comes to future predictions, and another error is applying our cultural views and values on past humans, which has yet to be clearly named as a bias. What can be clearly said about both spatial and temporal scales is that we are often irrationally biased towards very distant entities - in space or time - and even irrationally more than we should be. We are for instance inclined to reject the importance of a distant future scenario, although it may widely follow the same odds to become a reality than a close future. For example, almost everybody would like to win the lottery tomorrow rather than win the lottery in 20 years, irrespective of your chances to live and see it happen, or the longer time you may spend with your lottery prize for the (longer) time to come. Humans are most peculiar constructed beings, and we are notorious to act irrationally. This is equally true for spatial distance. We may care irrationally more for people that are close to us as compared to people that are very distant, even independent of joined experience (e.g with friends) or joined history (e.g. with family). Again, this infers a bias which we can be aware of, but which has to be named. No doubt the current social developments will increase our capacities to recognise our biases even more, as all these phenomena also affect scientists.

The following table categorizes different types of Bias as indicated in the Wikipedia entry on Bias according to two levels of the Design Criteria of Methods.

Type of Bias Category Description Relevant to research Qualitative Quantitative Individual System
Anchoring Cognitive Bias Anchoring' one's analysis on the first encountered piece of information (data) as a reference x x x x
Apophenia Cognitive Bias The tendency to perceive meaningful patterns within random data x x x x
Attribution Bias Cognitive Bias Systematic errors based on flawed perception of others' or one's own behavior (x) x x
Confirmation Bias Cognitive Bias The tendency to search for and favor information that confirms one's existent beliefs x x x x
Dunning Kruger Effect Cognitive Bias Lets lets lesser gifted people assume their superiority over others. (x) x x
Framing Cognitive Bias The way that individual actors present and construct evidence x x x x
Cultural Bias Cognitive Bias Interpreting and judging phenomena by standards inherent to one's own culture x x x
Halo / Horn Effect Cognitive Bias An observer's overall impression of an entity influences feelings about specifics of that entity's properties (x) x x
IKEA Effect Cognitive Bias Attributing a higher value to something one did by oneself x x x
Hindsight Bias Cognitive Bias Convincing yourself after an event that you knew it would happen all along x x x
Self-serving Bias Cognitive Bias The tendency to credit accomplishment to one's own capacities but failure to outside factors (x) x x
Status Quo Bias Cognitive Bias The emotional tendency to perceive any change to the current situation as deterioration x x
Bribery Conflicts of Interest Being compensated for an influenced behavior or opinion (x) x x
Favoritism Conflicts of Interest Favoring members of one's in-group over out-group members x x x
Lobbying Conflicts of Interest The act of influencing actors towards one's own interests (x) x x
Self-Regulation Issues Conflicts of Interest Inaccuracies occurring through self-versus independent evaluation x x x
Shilling Conflicts of Interest Pretending to be, but not being; independent; thereby deceiving observing individuals x x
Forecast Bias Statistical Bias Consistent differences between forecasts and the actual results x x x
Observer-expectancy Bias Statistical Bias The subconscious influence a researcher's expectations impose on the research x x x
Reporting Bias Statistical Bias Selective choice and publication of information, e.g. (un)desirable research results x x x
Social Desirability Bias Statistical Bias Survey respondents that tend to answer in a supposedly socially acceptable way x x x
Selection Bias Statistical Bias Unrepresentative Sampling x x x
Classism Prejudices Attitudes that benefit a specific social class (x) x x x x
Lookism Prejudices Prejudices based on physical properties, e.g. attractiveness or cultural preferences (x) x x x x
Racism Prejudices Behavior based on the assumption that there were inferior or superior (human) races x x x x x
Sexism Prejudices Behavior based on the assumption that one sex or gender (male, mostly) was better than others x x x x x
Academic Bias Biases in Academia Researchers who let their beliefs and world views shape their research x x x x
Experimenter Bias Biases in Academia Different experimenters assess subjective criteria differently, or individuals that are observed act differently when watched x x x x
Funding Bias Biases in Academia The tendency of a scientific study to support the interests of the study's financial sponsor x x x x x
Full-text-on-net Bias Biases in Academia Favoring open access journals in the references x x x
Publication Bias Biases in Academia Only publishing what fits in the narrative of the journal, or only results that are significant x x x x
Inductive Bias Other Biases Machine learning that predicts the future through algorithms based on specific training cases x x x
Other Biases Agenda Setting, Gatekeeping, Sensationalism, Educational Bias, Insider Trading, Implicit Bias, Match Fixing, Racial Profiling, Victim Blaming

Bias in gathering data, analysing data and interpreting data

The three steps of the application of a method are clearly worth investigating, as it allows us to dismantle at which stage we may inflict a bias into our application of a method. Gathering data is strongly associated with cognitive bias, yet also to statistical bias and partly even to some bias in academia. Bias associated to sampling can be linked to a subjective perspective as well as to systematic errors rooted in previous results. This can also affect the analysis of data, yet here one has to highlight that quantitative methods are less affected by a bias through analysis than qualitative methods. This is not a normative judgement, and can clearly be counter-measured by a sound documentation of the analytical steps. We should nevertheless not forget that there are even different assumptions about the steps of analysis in such an established field as statistics. Here, different schools of thought constantly clash regarding the optimal approach of analysis, sometimes even with different results. This exemplifies that methodological analysis can be quite normative, underlining the need for a critical perspective. This is also the case in qualitative methods, yet there it strongly depends on the specific methods, as these methods are more diverse. Concerning the interpretation of scientific results, the amount and diversity of biases is clearly the highest, or in other words, worst. While this is related to the cognition bias we have as individuals, it is also related to prejudices, bias in academia and statistical bias. Overall, we need to recognise that some methods are less associated to certain biases because they are more established concerning the norms of their application, while other methods are new and less tested by the academic community. When it comes to bias, there can be at least a weak effect that safety - although not diversity - concerning methods comes in numbers. More and diverse methods may offer new insights on biases, since one method may reveal a bias that another method cannot reveal. Methodological plurality may reduce bias. For a fully established method the understanding of its bias is often larger, because the number of times it has been applied is larger. This is especially but not always true for the analysis step, and in parts also for some methodological designs concerned with sampling. Clear documentation is however key to make bias more visible among the three stages.

Bias and philosophy

The last and by far most complex point is the root theories associated to bias. Reason, social contract and utilitarianism are the three key theories of Western philosophy relevant for empiricism, and all biases can be at least associated to one of these three foundational theories. Many cognitive bias are linked to reason or unreasonable behaviour. Much of bias relates to prejudices and society can be linked to the wide field of social contract. Lastly, some bias is clearly associated with utilitarianism. Surprisingly, utilitarianism is associated to a low amount of bias, yet it should be noted that the problem of causality within economical analysis is still up for debate. Much of economic management is rooted in correlative understandings, which are often mistaken for clear-cut causal relations. Psychology also clearly illustrates that investigating a bias is different from unconsciously inferring a bias into your research. Consciousness of bias is the basis for its recognition: if you are not aware of bias, you cannot take it into account regarding your knowledge production. While it thus seems not directly helpful to associate empirical research and its biases to the three general foundational theories of philosophy - reason, social contract and utilitatrianism -, we should still take this into account, least of all at it leads us to one of the most important developments of the 20th century: Critical Theory.

Critical Theory and Bias

Out of the growing empiricism of the enlightenment there grew a concern which we came to call Critical Theory. At the heart of critical theory is the focus on critiquing and changing society as a whole, in contrast to only observing or explaining it. Originating in Marx, Critical Theory consists of a clear distancing from previous theories in philosophy - or associated with the social - that try to understand or explain. By embedding society in its historical context (Horkheimer) and by focussing on a continuous and interchanging critique (Benjamin) Critical Theory is a first and bold step towards a more holistic perspective in science. Remembering the Greeks and also some Eastern thinkers, one could say it is the first step back to a holistic thinking. From a methodological perspective, Critical Theory is radical because it seeks to distinguish itself not only from previously existing philosophy, but more importantly from the widely dominating empiricism, and its societal as well as scientific consequences. A Critical Theory should thus be explanatory, practical and normative, and what makes it more challenging, it needs to be all these three things combined (Horkheimer). Through Habermas, Critical Theory got an embedding in democracy, yet with a critical view of what we could understand as globalisation and its complex realities. The reflexive empowerment of the individual is as much as a clear goal as one would expect, also because of the normative link to the political.

Critical Theory is thus a vital step towards a wider integration of diverse philosophies, but also from a methodological standpoint it is essential since it allowed for the emergence of a true and holistic critique of everything empirical. While this may be valued as an attack, I would value it as a necessary step, since the arrogance and the claim of truth in empiricism can be interpreted not only as a deep danger to methods. Popper does not offer a true solution to positivism, and indeed he was very much hyped by many. His thought that the holy grail of knowledge can ultimately be never truly reached also generates certain problems. He can still be admired because he called for scientists to be radical, while acknowledging that most scientists are not radical. In addition, we could see it from a post-modernist perspective as a necessary step to prevent an influence of empiricism that might pose a threat to and by humankind itself, may it be through nuclear destruction, the unachievable and feeble goal of a growth economy (my wording), the naive and technocratic hoax of the eco modernists (also my wording) or any other paradigm that is short-sighted or naive. In other words, we look at the postmodern.

Critical Theory to this end is now developing to connect to other facets of the discourse, and some may argue that its focus onto the social science can be seen critical in itself, or at least as a normative choice that is clearly anthropocentric, has a problematic relationship with the empirical, and has mixed relations with its diverse offspring that includes gender research, critique of globalisation, and many other normative domains that are increasingly explored today. Building on the three worlds of Popper (the physical world, the mind world, human knowledge), we should note another possibility, that is Critical Realism. Roy Bhaskar proposed three ontological domains (strata of knowledge): the real (which is everything there is), the actual (everything we can grasp), and the empirical (everything we can observe). During the last decade, humankind unlocked ever more strata of knowledge, hence much of the actual became empirical to us. We have to acknowledge that some strata of knowledge are hard to relate, or may even be unrelatable, which has consequences for our methodological understanding of the world. Some methods may unlock some strata of knowledge but not others. Some may be specific, some vague. And some may only unlock new strata based on a novel combinations. What is most relevant to this end is however, that we might look for causal links, but need to be critical that new strata of knowledge may make them obsolete. Consequently, there are no universal laws that we can thrive for, but instead endless strata to explore.

Coming back to bias, Critical Theory seems as an antidote to bias, and some may argue Critical Realism even more so, as it combines the criticality with a certain humbleness necessary when exploring the empirical and causal. The explanatory characteristic allowed by Critical Realism might be good enough for the pragmatist, the practical may speak to the modern engagement of science with and for society, and the normative is aware of – well - all things normative, including the critical. Hence a door was opened to a new mode of science, focussing on the situation and locatedness of research within the world. This was surely a head start with Kant, who opened the globe to the world of methods. There is however a critical link in Habermas, who highlighted the duality - if I may - of the rational individual on a small scale and the role of global societies as part of the economy (Habermas 1987). This underlines a crucial link to the original three foundational theories in philosophy, albeit in a dramatic and focused interpretation of modernity. Habermas himself was well aware of the tensions between these two approaches – the critical and the empirical -, yet we owe it to Critical Theory and its continuations that a practical and reflexive knowledge production can be conducted within deeply normative systems such as modern democracies.

Linking to the historical development of methods, we can thus clearly claim that Critical Theory (and Critical Realism) opened a new domain or mode of thinking, and its impact can be widely felt way beyond the social science and philosophy that it affected directly. However, coming back to bias, the answer to an almost universal rejection of empiricism will not be followed here. Instead, we need to come back to the three foundational theories of philosophy, and need to acknowledge that reason, social contract and utilitarianism are the foundation of the first empirical disciplines that are at their core normative (e.g. psychology, social and political science, and economics). Since bias can be partly related to these three theories, and consequentially to specific empirical disciplines, we need to recognise that there is an overarching methodological bias. This methodological bias has a signature rooted in specific design criteria, which are in turn related to specific disciplines. Consequently, this methodological bias is a disciplinary bias - even more so, since methods may be shared among scientific disciplines, but most disciplines claim either priority or superiority when it comes to the ownership of a method.

The disciplinary bias of modern science thus creates a deeply normative methodological bias, which some disciplines may try to take into account yet others clearly not. In other words, the dogmatic selection of methods within disciplines has the potential to create deep flaws in empirical research, and we need to be aware and reflexive about this. The largest bias concerning methods is the choice of methods per se. A critical perspective is thus not only of relevance from a perspective of societal responsibility, but equally from a view on the empirical. Clear documentation and reproducibility of research are important but limited stepping stones in a critique of the methodological. This cannot replace a critical perspective, but only amends it. Empirical knowledge will only look at parts - or strata according to Roy Bhaskar - of reality, yet philosophy can offer a generalisable perspective or theory, and Critical Theory, Critical Realism as well as other current developments of philosophy can be seen as a thriving towards an integrated and holistic philosophy of science, which may ultimately link to an overaching theory of ethics (Parfit). If the empirical and the critical inform us, then both a philosophy of science and ethics may tell us how we may act based on our perceptions of reality.

Further Information

Some words on Critical Theory
A short entry on critical realism

The author of this entry is Henrik von Wehrden.