How do we develop scientific knowledge?

Last month I wrote an article criticising the NHS for offering homeopathy, a treatment that has no value according to medical science. A guest writer for TheKnowledge responded by saying homeopathy is “beyond the understanding of science” and even accusing researchers debunking it of fabricating results- pointing to the conflicts of interest that can occur when big pharma is funding research. So which is it Mr Ives: is homeopathy beyond the understanding of science or is the science being faked in order to support a malicious agenda?

Concerning the latter claim I would like to make absolutely clear that I am aware of the massive problems arising from conflicts of interest within pharmacology. Taking an example which is close to my heart, that I have been trying to raise awareness of in my personal and academic life, is the corruption of psychopharmacology by big pharma. This topic is something I would like to write about in more depth but suffice to say, there is mounting evidence to support the idea that antidepressants don’t work (See “The Emperor’s New Drugs: Exploding The Antidepressant Myth” by Dr Irving Kirsch for an easy to read argument without too much jargon, I will include references to peer-reviewed journals in the full article to be written in future). Obviously, this is a touchy subject for many with anecdotal evidence, just like those with evidence supporting the efficacy of homeopathic medicines, but we must keep in mind that we are all susceptible to biases that inhibit our ability to remain objective and impartial.

Returning to the claim about homeopathy being “beyond the understanding of science”, I first would like to repeat how disappointed I am that a university student would be so scientifically illiterate. His ignorance, however, has inspired me to write a bit on what science is. I’ll also take this opportunity to say that his self-deprecating jibe about Christianity, “[…]never have I been accused of believing in something that has no scientific evidence. Unless you count that I’m a Christian”, shows his ignorance again. It infers that one cannot be simultaneously a scientist and a Christian, which is not the case. Further, the atheism debate is actually beyond the scope of science and something I’d like to discuss in future. Suffice to say, I think the scientist’s position should be one of skepticism and therefore absurdism- a confession that we simply do not know and we may never know.

Below is an edited version of an old essay I wrote at university, quite a while ago. I’ve removed references that were specific to the essay as it was written about social science. If you find yourself in disagreement or simply want more detail, any introductory textbook should be satisfactory or feel free to contact me.

Scientific knowledge is the cumulative, systematically acquired collection of theories, hypotheses and data gained through the scientific method of inquiry. This method organises and increases knowledge with the aim of furthering our understanding of the universe so that we can explain natural phenomena; as well as enabling us to make more accurate predictions about possible future changes in the world. Experiments are fundamental to the overall research process which includes systematic empirical observation- the purpose of which is to form, develop and falsify hypotheses. Moreover, they are an amalgamation of different schools of thought, which in a research environment, allow us to continuously create and/or edit existing theories that are more valid and reliable. Due to this, replication and falsification are a crucial feedback loop in the development of scientific knowledge.

Experiments are important for creating scientific theories as they allow researchers to purposefully limit or control certain variables so that we may deduce more valid conclusions from the data gathered. This action itself however, may have an effect on the results of experiments which is why some researchers prefer the method of naturalistic observation (observing natural phenomena as it occurs in the world without any intervention or assistance from the researchers). Even the initial data gathering as part of an experiment may have a significant impact on its results- this is known as selection bias. For example, a researcher enquiring into the relationship between drug effects versus drug dosage may discover different results depending on the age of the participants in their sample- older participants may feel larger drug effects per dosage compared to younger participants.

Another factor for consideration to ensure maximal validity and reliability of results is time. Longitudinal studies are quasi-experimental research designs in which data is studied and compared with itself as time passes. For example, a researcher enquiring into the effects of how differing environments effect a pair of twins over time would simply gather data at the start of the experiment then gather data in the same manner later on, without actually controlling or manipulating any of the variables which could have an effect on what they are studying. Unfortunately, using this research design severely restricts the amount of valid conclusions we can deduce from the results/data, as it will have extremely low internal validity. This means that even if we notice certain correlations or patterns in results/data, we cannot deduce real causal links.

In order for us to be able to deduce actual causal relationships between variables studied, researchers must design and implement True experiments. There are two main types of True experimental designs (in social sciences): between-subjects designs and within-subjects designs. A between-subjects design compares groups of data so that researchers may deduce causal relationships between them. This comparison is possible because researchers control or manipulate some variable(s) in order to separate the sample into groups that can be compared. Although this is good for the internal validity of the experiment as we can more accurately deduce relationships between groups of data, the external validity or replicability and reliability of the experiment is simultaneously damaged. For example, a researcher enquiring into the relationship between mathematical ability scores between students studying different courses, may find business students score higher than geography students on this score. However, this does not give the researcher a valid enough argument to soundly conclude that geography students, in general, have less mathematical ability than business students because external factors, such as where students are studying, may have impacted the results of the experiment, at the point during which the sample was taken.

On the other hand, within-subjects experimental designs have great external validity and replicability to the detriment of internal validity. Within-subjects designs involve gathering data after manipulating some variable(s) in order to see how these manipulations produce differences within the overall sample. In other words, the sample is subject to differences within itself via manipulations of this variable, as opposed to the previous design, in which the sample is divided into groups before the data is gathered. In the former, this allows us to see individual differences within the overall sample, between the groups created by the experiment; in the latter, this allows us to see individual differences between the groups created before the experiment starts. One example of a within-subjects design could be a researcher enquiring into the relationship between efficacies of different therapeutic treatments. In this case, the researcher would simply apply each treatment to the entire sample consecutively and later look for correlations or patterns that separate individuals in order to form comparable groups.

The formation of comparable groups allows us to analyse the differences between them in order to form theories about potential causal relationships. In this stage of the process of acquiring scientific knowledge, logic and the rational method is employed. The most useful idea about how to deduce arguments from the data gathered is empirical falsification. Empirical falsification helps scientists work around the problem of induction. This problem is simply put “how is it possible to claim knowledge about future instances based on knowledge derived from similar instances that have already occurred?”. Basically, if we have a hypothesis about some phenomena, we can investigate and find certain laws about those phenomena but we can never be certain that these laws will hold true in future, we can only say it is probable, to some extent, that the phenomena will follow these laws again in future, considering we have no evidence that would suggest otherwise. If we discover evidence to the contrary, then the hypothesis is falsified by the new data meaning we must append, edit or completely remake our theories about the laws governing those phenomena.

To truly allow us to make realistic general conclusions about the specific results observed in experiments, analysis must be done using statistical procedures. However, statistical procedures come with their own set of problems and biases which can impact the results of experiments. Selection bias however, can be reduced by random sampling, as this will produce a sample that is more representative of the study population.

Selection bias means that a sample can never truly be 100% representative of the study population as the sample can only ever represent itself fully. To convey the possible error inherent in this potential difference in representativeness, scientists employ confidence margins as a way of illustrating how accurate a theory represents the study population. Essentially, this means that the confidence margin is the probability that the theory accurately describes any sample from the original population. For example, a research project concludes that all post-graduate mathematics students have completed undergraduate courses of the same subject, with a confidence margin of 95%. This would mean that any other researcher should be able to take a sample of post-graduate mathematics students and be 95% confident that the sample they choose has completed undergraduate courses of the same subject.

Modern ethical practice of scientific research requires the exercise of informed consent- meaning participants in studies should be made aware that they are participating in a study so they are not left feeling deceived by the researchers later. In social scientific research, informed consent causes selection bias, due to participants volunteering to participate in studies, being a factor that may potentially affect results. Truly random sampling would not provide informed consent but is considered unethical. The problem of selection bias has led some scientists to wrongly believe that non-informed consent is ethical whereas it is a breach of the ethical principle of beneficence.

In the past, some research has been carried out without participant’s informed consent or awareness. This has been allowed because collected data has been anonymised and stored securely. However, anonymity does not guarantee ethical approval because another important factor for consideration during the ethical review process is confidentiality. Confidentiality is the right of the participant to decide who can access the data produced as part of an experimental investigation. Research ethics is another topic I would like to return to in another article so please forgive this section for being shallow.

Overall, the development of scientific knowledge is a continuous process that answers questions using testable hypotheses so that we may make realistic deductions about the true nature of the observable world. This method incorporates all other methods of enquiry in a manner which is sustainable, holistic and idealistically ethical. The key difference between each separate method and the overall scientific method is it is adaptable depending on the circumstances dictated by current social norms. Social science specifically, is a unifying, naturalistic, collaborative and interpretive approach to asking questions about the nature of social reality, one could call it the study of “social physics”.

As you can see Mr Ives, scientists aren’t lacking in self-awareness. In fact, peer review, proof-reading, replication, falsification etc are among the key reasons as to why science has been doing so well for the past few centuries and show how seriously we take our endeavours. Yes admittedly there are those that abuse science to support their agenda but that is why we should actually investigate shady looking evidence instead of just making assumptions. Sometimes through scientific pursuits we discover things about the world we might not want to be, nevertheless scientists are morally obligated to report the truth, no matter how difficult it can be to accept.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s