.png){kind=icon}
Ethics is a theme running throughout our book, and it is worth emphasizing that you should always have ethical considerations at the forefront of your mind when designing and conducting experiments - and never more so than when you are dealing with human subjects. No matter how trivial you consider the involvement of the subjects in your study, you should take time to think about how being a subject in your experiment could impact on an individual’s self-regard and dignity. You must do everything you can to minimize any potential negative impacts.
Put yourself in the position of the subject. They will generally be in unfamiliar surroundings being asked to do things that they do not normally do. As the person instructing them, you are likely to be perceived as an authority figure. It is in your power to make the subjects feel as comfortable and relaxed as possible. It is very easy for subjects to feel that your data collection is a test in which their self-respect makes them want to perform as well as possible. Although they are just, for example, anonymous 20-year-old females in your study, they can feel as if they personally are being tested. It is very easy for them to lose self-esteem after feeling that they performed less than perfectly. It’s important that you do all that you can to avoid this feeling.
One important issue with humans is the idea of informed consent: that not only does the person agree to be part of your study but they give their consent in the knowledge of why these data are valuable and how they are going to be used. As much as is possible, the subjects should have forewarning of what is going to happen to them and what will be asked of them. Volunteers should also be fully aware that they are entitled to stop participating at any point and withdraw from the study without having to justify this decision to anyone. This may seem like a nuisance to you, and it is, but you must remember that high levels of drop-outs of this kind are likely to be caused by failing to give proper forewarning of demands made on the subjects in your initial briefing of volunteers whilst you were collecting their informed consent. You will get the odd person who withdraws half way through the data collection for reasons that are impossible for you to imagine or understand. Don’t stress about this; if you have designed your data collection properly then the loss of this individual should not jeopardize your experiment, and you should take heart that you have created the appropriate atmosphere where people are comfortable acting on their right to withdraw rather than suffering in silence. Similarly, you will get the odd person who does not seem willing or able to cooperate in data collection as you expected. If this happens frequently, then there is something wrong with your data collection techniques, but if it is just a very rare individual, then don’t get frustrated, just chalk it up as an interesting experience that reminds you of the tremendous variety of human personality. No matter what you think, treat that person with dignity and professionalism, collect the data from them as best you can, send them on their way without any feeling that you’ve been inconvenienced, and think at the end of the day (when you can look back on the incident with a broader perspective) whether you should discard that individual from your data set or not.
No matter how fully people were briefed beforehand, you should always make sure that your experimental schedule does not have the subjects bundled back out onto the street within 15 seconds of you collecting the data from them. Someone friendly and knowledgeable should be made available to them afterwards to answer any questions or concerns that arose during the data collection. Not all subjects will want to avail themselves of this opportunity, but it should be available to those that do. There will often be aspects to the work that you have to hide from the subjects before the data collection so as not to influence their performance or answers to questions, but that information should be made available to those that are interested after you have collected their data. This brings us on to the important concept of deception.
Deception
Sometimes the logical way to get the most accurate information from a subject would be to deceive them. Imagine that you want to simply measure the prevalence of left-handed writing amongst people in the local population. You feel that the best way to do this is to ask volunteers to sit at a table and fill out a short form with a pen. Both the form and pen are placed centrally and symmetrically on the table directly in front of where the subject is asked to sit. You then note down whether the person uses their right or left hand. You are collecting data by deception. Is this ethical? Scientists have different standpoints on this. Some would say that there are no circumstances where collecting information by deception is appropriate, and this is a viewpoint that we consider to be entirely valid. Our viewpoint is that there are some circumstances where addressing a valid scientific question can only be practically addressed by collecting the data by deception. If you can demonstrate both that your question is worth asking, and that it can only be addressed by deception, then such deception may be appropriate.
Returning to our example, we would consider that measuring the local prevalence of left-handed writing could well be a scientifically valid thing to measure, but we are not convinced that these data could only be collected by deception. Firstly, we are not convinced that simply asking people whether they write with their right or left hand (or both) would not be effective. It seems unlikely that people would be unaware of this, and we can see little motivation to be dishonest in their answer. However, if you were concerned about dishonesty, then perhaps you could collect data on buses and trains by watching people fill in crossword puzzles in their newspapers. Thus, in this case (and the vast majority of cases) deception cannot be justified. Even if deception can be justified, we would strongly recommend that in almost all circumstances you then debrief your subjects, explaining the nature of the deception, the purpose of the data collection, and why you felt that deception was necessary to collect the data. You should then ask them if they are comfortable with the data being used for the stated purpose, and explain that you’d be entirely prepared to strike their data from your study if they feel at all uncomfortable with your use of deception. If they do feel such discomfort, then you should apologize, explaining your motive again if appropriate.
Collecting data without permission
In the last section we suggested that you could collect data on handedness by observing people filling out crossword puzzles in their newspapers in public places like buses and trains. Is this ethical, in that you are collecting data from people without their knowledge? We would be comfortable with data collected in this way provided the data include nothing that would allow specific individuals to be identified if someone were to get hold of your notes, and providing that the information is in the public domain. By this we mean that their behaviour or other measure is something that the subject is comfortable with other people observing. If you fill out a crossword on a train then you can be presumed to be unconcerned with allowing others to observe you in this act. Thus, collecting data on aspects of this should not be considered intrusive. But looking through people’s dustbins to see what foodstuffs they eat or peering into kitchen windows to record what time of day meals are prepared is unduly obtrusive.
You should also take care that data collected in this way do not involve you in anything that would make people feel uncomfortable. Say you want to know how many shops the average shopper visits in 30 minutes. Picking people at random and following them around your local shopping centre for 30 minutes would be irresponsible and unethical. Imagine how you would feel if you noticed a complete stranger tailing you around the shops. We’d phone the police in such a situation.
Confidentiality
There are very few experiments where it is necessary to record someone’s name. If you don’t need their name then don’t record it. There may be circumstances where you do need to record personal information such as names and telephone numbers, perhaps if your study involves repeated follow-ups. You must be very careful with such data, never allowing it to fall into other hands. It would also be wise to avoid keeping such data in the same place as sensitive information about individuals. For example, if you require individuals to fill in personality questionnaires at 6-month intervals, you should give every subject a code number (or pseudonym), and this code rather than their name should be used in your files that contain the personality information. Have a separate file, kept quite apart from the personality information, that holds the name and contact details of individuals along with their code numbers. This may seem like unnecessary hassle for you, but think how you’d feel if someone else’s carelessness allowed your personal details to fall into the wrong hands. So treat confidentiality very seriously, and take time to explain to your subjects how seriously you take confidentiality.
Discretion
Sometimes you must take sensitive information from people, and you must go about this in a sensitive way. We have no problem with asking people in the street whether they are left- or right-handed, as this question does not require a lot of thought to answer and is not likely to cause discomfort to the person questioned. However, asking someone about the number of sexual partners they have had in their lifetime or their views on the ethics of eating meat does not fall into this category. Hence, these are not appropriate questions to ask someone you’ve stopped in the street. You will get more accurate data and protect your subjects better if you ask them these questions in a more appropriate environment. For example, you might still stop people in the street and explain that you are surveying people’s opinions on the ethical implications of what we eat, asking them if they would be comfortable filling out a questionnaire for you in their own time and posting it back to you in the freepost envelope provided. Or you might ask them if they would take away a flier that explains the purpose of the survey and invites those interested in taking part to phone a certain number to arrange a time when they are willing to come to your institute to be interviewed. Now each of these solutions has problems of introduced bias, because there is the potential that a non-random sample of people asked will respond, but by careful design of your questionnaire or interview, you can evaluate the extent of this bias. For example, you could collect data on the age of respondents and compare this to existing data on the age-structure of your statistical population to explore whether young people, for example, seem more willing to participate. You can then combat any highlighted deficiencies in your sample by targeting certain types of people in subsequent recruiting sessions. In contrast to this, asking people questions on complex or sensitive issues in the street is certain to produce poor-quality data and give you no real way to evaluate how poor they are.
Randomized response
Sometimes you might find that you need to gather information of a sensitive or delicate nature. In such a situation, simply asking people directly is unlikely to be successful, as individuals may often provide inaccurate answers. Randomized response is a method of making it easier for human subjects to give embarrassing or incriminating answers to questions. For example, if surveying drug use, it is likely that some interviewees will lie and deny illegal drug use. In order to at least partially overcome this, you might conduct a telephone interview along the following lines:
“I'm now going to ask you a potentially embarrassing question and this is why I asked you to have a die ready. I want you to roll the die now, before I ask you the question. If you roll a six, then answer my question with “yes” no matter whether that is the true answer to the question I'm about to ask or not. If you roll another number, then answer the question truthfully. This means that neither I nor anyone other than yourself will know if you answered truthfully or not, but if we ask this question to a lot of people we can work out the average fraction of truthful responses. Now, here is my question: have you used illegal drugs in the last six months?”
If we ask 600 people this question and 150 answered yes, what can we conclude? We expect that on average 100 people will have rolled a six and answered yes, so on average 500 people will have rolled another number of which 50 answered “yes”, so we estimate that 10% of the surveyed people have used illegal drugs in the last six months.
A small drawback to this technique is that sample size has to be increased because a fraction of people, 1/6 in this case, are instructed not to answer honestly. It also requires greater organization from both the interviewer and interviewee. However, this may be worthwhile if it ensures greater honesty from participants. Another issue to consider is to emphasize to subjects that their answer is useful to the study even if they roll a 6 and don’t give a true answer to the question. This can be dealt with by careful phrasing of the instructions regarding answering, in a way that also explains how this approach allows collection of population level statistics without recording sensitive information about individuals.
Ethical guidelines
The institution under whose auspices you are carrying out experiments is almost certain to have a set of ethical guidelines and/or an ethics committee that has to approve experiments that involve humans. If you are carrying out the experiment then it is your responsibility to make sure that you are complying with the local ethical requirements. Do so in plenty of time before the start of the experiment, in case your particular experiment needs approval by a local ethical committee, since such committees may sit only infrequently. If your institution does have ethical guidelines then it almost certainly has sanctions to impose on those that do not comply.
If in doubt, you should, of course, seek advice from others about the institution’s policies. But make sure you get authoritative advice. If you are carrying out the experiment then the onus is on you to have the correct permissions.
If you are planning to publish your work then journals too often have published sets of ethical guidelines. It is worth being familiar with those of likely target journals for your work.
Volunteers
Volunteers are great! It is amazing what good-hearted people are willing to put up with (often for no financial reward) to help someone out with a scientific study. Science would be in a lot of trouble without volunteers. But take time to think what population you are looking to sample from and whether your volunteers are an unbiased sample from that population. Imagine that you are stopping people in the street and asking them whether they are right- or left-handed. If the population that you are interested in are adults in your local area, then you would doubtless be careful to collect samples from different streets and at different times of day so as to get as representative a sample as possible. You’d also be careful to make sure that you are unbiased in your approaches to people. This can easily occur, as it is perfectly normal to feel more comfortable approaching people of your own age group, for example, in such situations. Even if you avoid all these potential sources of bias in your sampling, there is always the danger of bias occurring because not all people approached will volunteer to answer your question. If, for example, handedness was more strongly associated with personality traits like openness and helpfulness, then your data set could be biased because one group (right- or left-handers) would be more likely to participate in your study than the other. You should weigh up the likelihood of such biases being a problem in any study using volunteers, and if there is a reasonable likelihood of such a problem in your study, then redesign your experiment to avoid this problem or to allow you to evaluate whether such a bias was a significant factor in your data collection.
Honesty of subjects
When your data collection involves asking people questions, you have got to consider the possibility that people may be dishonest in their answer. This can occur for a number of reasons. It may be that they are trying to give you the answer that they think you want to hear. It could be that they are seeking to impress you, or that they find the honest answer embarrassing. There are a number of things you can do to combat this. Most importantly, you can create an atmosphere in which the subject understands the need for honesty, and feels that your consideration of confidentiality and discretion will avoid any need to lie for reasons of potential embarrassment. This atmosphere is created by your initial discussions when you recruit the subjects and obtain informed consent. It is not just about what you say, but how it is said and by whom. Maintain a friendly but professional atmosphere. It is best if the questioner is seen to be disinterested in the subject except as the subject of a scientific study. That is, if you have some relationship to the answerer (e.g. you are a colleague, or a healthcare professional involved in the subject’s treatment), then it’s probably better if you get someone else to conduct the interview. Explicitly warn subjects about giving the answers that they think you want to hear, and aid this process by avoiding asking leading questions. Lastly, if you are gathering information about issues where dishonesty is likely (e.g. illegal activity or sexual behaviour) then it may be wise to ask a series of questions that effectively collect the same information, allowing you to check for inconsistency in responses, as indicators of dishonesty. You might choose to (tactfully) confront the subject with this apparent inconsistency during the interview in the hope of getting to the truth, or delete such inconsistent individuals from your sample.