This is how scientists go about observing, analyzing the results of their observations, and arriving at predictions.  The key thing about this method is that it is designed to arrive at valid conclusions in spite of human fallibility. 

We do not propose to discuss all aspects of the Scientific Method.  There are any number of online sources that do that quite well (see below).  Rather, we wish to address some key aspects of the Scientific Method that deal with many of the criticism raised toward science.

General Outline: This is a brief look at how the Scientific Method works

The scientific method is the best way yet discovered for finding the truth and differentiating it from lies and delusion. The simple version looks something like this:

1. Observe some aspect of the universe.

2. Invent a tentative description, called a hypothesis, that is consistent with what you have observed.

3. Use the hypothesis to make predictions.

4. Test those predictions by experiments or further observations and modify the hypothesis in the light of your results.

5. Repeat steps 3 and 4 until there are no discrepancies between theory and experiment and/or observation.

When consistency is obtained the hypothesis becomes a theory and provides a coherent set of propositions which explain a class of phenomena. A theory is then a framework within which observations are explained and predictions are made.

Empirical:

As noted above, the scientific method involves empirical observation of observable facts.

Prediction:

The scientific method involves predicting outcomes based on cause and effect relationships.  This is a key aspect of validity in science.

Experimental studies:

This is the means for testing predicted outcomes, usually in the laboratory.  Experiments are the surest test for causality.  There are a number of different experimental designs. 

Perhaps the simplest experiment involves establishing a condition, or situation (e.g., twenty potted plants), dividing that into two parts, the experimental group and the control group (ten potted plants in each group) and apply the experimental condition to one group (water ten potted plants and not water the other ten potted plants) and observe the result (the experimental group lives, the control group dies.

Non-experimental studies:

Sometimes it's not possible to conduct a laboratory experiment, so other forms of research have been developed.  While they suggest causal relationships, they do not prove them.  These often rely on increasingly sophisticated forms of correlational data analysis (e.g. multiple regression and discriminate function analysis).

Correlational study: exams how one variable changes when another variable changes.  While it does not demonstrate causality, it suggests there is some sort of relationship between the two variables.  Here are three graphs showing a hypothetical correlational relationship between to variables concerning automobiles.  In the negative correlation, the more mileage on the car the lower the value (when one variable goes up, the other goes down), while in the positive correlation, the few accidents the lower the cost of insurance (when one variable goes up, so does the other), and a zero correlation occurs when body color changes and auto quality changes (when one variable changes, the other doesn't change).

Published methods, data analysis and results: 

The key here is the science is a scholarly enterprise, which means that it is based on shared information, which is done by means of publications in scientific journals, and text books.  This was a major advance in how humans came to understand the world we live in.  As a result, scientists are required to share all aspects of their research, including their source of data, methods of collecting data, data analysis and results, all of which provides for...

Peer review and replication (a group effort):

Scientific findings, including the validity of the research methodology and the potential application beyond the experiment itself, are reviewed by other scientists.  In other words, multiple scientists are involved in assessing published research findings.  There are two main steps in this process.

Peer review occurs when research is submitted for publication.  The article is sent to other researchers who assess the methodology and findings and then report whether they think the research merits publication.  This does not guarantee that the research is fully valid.  It simply says it is worth sharing.  That is a key element of science, the sharing of information so others can also examine and work with it.

It is important to note that not all science publications are peer reviewed, and scientists know which ones are and the level of review applied by journals.  In the case of those that don't use journal review, articles are published in order to make them public and allow for follow-up research.

Replication involves further repeated observation and experimentation (i.e. follow-up research), often by more than one scientist,  is required to validate the observation and the findings (you may recall that some years ago, some scientists ran an experiment that produced cold fusion, a highly desirable possible energy source, but no one could replicate their results).

Three levels of acceptance:

Scientific Method works with hypotheses, models and theories (laws of nature).  A hypothesis is something used to look at specific predicted research outcomes.  A model is used when it is known there are probable limitations to validity but a working framework is necessary for communication between scientists and to help direct ongoing research.  A theory, or law, contains a hypothesis (or a number of hypotheses) that have been confirmed.

Scientific consensus:

First, this does NOT involve agreement of opinion, or belief.  What it does involve is the three levels of acceptance listed above.  Scientists scrutinize each other, arrive at agreement about which hypotheses have merit and utility, which models result in empirically sound research (whether causal, or correlational).  And, finally, which theories, or laws can be supported and shared (e.g., the law of gravity).

Growth and change:

Of course, over time, given the goals and methods of science, things (including the consensus) change.  The body of scientific knowledge is constantly growing and changing (e.g., the previous models of the atom have been significantly altered as more research has been conducted) At the same research methods have grown, as has data analysis techniques (more use of fuzzy logic and probability).  Einstein's theory of relativity expanded our view of physics beyond earlier laws of Newton.

Faulty, or Invalid Research:

Yes, this can happen.  The Scientific Method makes it more difficult, but it can't prevent all errors, or malfeasance.  Nothing can do that.  For the most part, there are two ways for invalid research to get published, and in most cases it will be rectified (because some scientist will figure it out).

Unfortunately, many of current criticisms of science are directed at the scientists themselves (e.g., "They're just in it for the money").  When these lack any review of the research, itself, these are simply ad hominem attacks (see our discussion of logical fallacies for more about this), and thus dismissible.

But scientists are human and thus fallible, so they do make mistakes, either on purpose, or by accident.

Intentional: Well, yes there are unscrupulous and malfeasant folks that sometimes end up in the scientific community.

A primary form of intentional invalidity is simply faking the data (making it up).  This has happened.  We often end up knowing that it is faked, either because no one has been able to replicate the data, or someone blew the whistle, or both.

Another example of intentional support of invalid research has been faked peer review.  Over time, these become apparent, and the fact that so many such fake reviews (usually in limited topic areas) have come to light verifies the extent to which scientists oversee each other.

Unintentional: Again, scientists make mistakes, no matter how hard they try. 

Methodological errors:

This can involve something as simple as malfunctioning measurement equipment to the wrong choice of subjects for an experiment. 

For example, many years ago, a scientist studying attention used cats in his study to see how they responded to stimuli, specifically sound localization.  His results (which were replicated) found cats didn't orient well to sounds from behind.  So everyone accepted this, until someone realized cats turn their ears, and if they couldn't hear what was behind, because their ears were turned in the wrong direction, then they wouldn't respond to sounds behind (unless they were quite loud).  He accidentally chose the wrong subjects for his study, and his results were valid for cats but but not because of the cognitive (brain oriented) process of attention, and they could not be generalized to other organisms.

Biases:

Scientists have biases, just like everyone, but they are supposed to occur in places where they have the least chance of producing invalid results.  So, scientists get to choose the things they want to study (that's a bias).  In the case of the study above using cats, it was the experimenters bias that led to using them (they were easily available and known to be pretty stimuli responsive). 

The primary bias scientists are expected to avoid is confirmation bias, that is trying to prove something they already believe to be true by designing their study to confirm what they want to find. 

Scientists routinely seek to confirm, or disconfirm an hypothesized result from their research, but they are supposed to do the research in a way to allow any and all possible results.  A classic example of disconfirming results occurred some years ago, when psychologists set out to demonstrate that depressed people tend to see things more negatively than they really are.  They found out they were wrong, that in some cases (particularly those where there was a risk of a highly negative event occurring)  people who were depressed saw reality better than non-depressed individuals, which led them to seeing the depressed folks as "sadder but wiser" (if you do an online search of this "sadder but wiser" you can read about this research).

Confirmation bias can and often does lead to biases in research design (e.g., choice of data to collect, from what source and how to analyze it), which is often caught in peer review.  Sometimes errors in methodology produce valid results for the study alone (called internal validity) but can't be generalized (external validity).  This can particularly happen when a sample is taken from the wrong source (as in the study of cats mentioned above), or when disconfirming date is ignored. 

A classic example of this was research conducted some years back to prove that gun ownership prevents crime.  The study concluded lower crime rate in a town that had more people armed with guns, in comparison to a similar town with less gun ownership.  The researcher failed to collect data from high crime, often inner city communities, where there is a very high rate of gun ownership and possession, as well as above average crime rates (especially homicide).  The study results were valid internally but could not be generalized to all communities due to sampling error.

Finally, there is a kind of bias known of as a "paradigm," which is a broad philosophical and/or conceptual framework.  An example would be the notion that the Earth is the center of the universe. 

Scientists are generally aware of this issue and the shifts in paradigms that have occurred over time, so, while they may operate under a given paradigm (complex computations were provided to explain the movement of the planets when it was thought that the Earth was the center of the universe, complex computations that worked under that paradigm), but revolutionary thinkers and researchers help change the paradigms (e.g., science moved from Newtonian physics, to relativity and quantum mechanics).

Self Correcting:

Science arose as a means for dealing with human fallibility in the quest for understanding and predicting the world around us.  As discussed above, by making it such a group effort, it created ongoing self-scrutiny.  The result has been research that has been withdrawn from publication.  Changes in peer review methods.  Ongoing paradigm shifts.

  In this case, two words: Science and Technology

Science can lead to technology, but they are not the same.  This is an important distinction.  Technology, or the way we do things, predates science by many millennia.  Technology is the way we use things to manipulate and change the world around us.  It dates all the way back to the earliest human tools.  But as science, a rather recent activity, has developed, it has led to more and more sophisticated technology (e.g., tools), such as light bulbs.  Early technology relied on trial-and-error, but the advent of systematic science led to much more rapid technological development.

 For a YouTube description of the Scientific Method by Physicist Richard Feynman

 For an online search for further information about the Scientific Method 

 For an extended discussion of the Scientific Method available on the internet

 To go to the introduction to science