Elliott talks about the value-laden decisions that scientists (and in this specific paper, nanotoxicological scientists) have to make on a regular basis, showing that science is not free of non-epistemological values.
Scientists incorporate epistemic or cognitive values in their judgements. Epistemic values promote the truth-seeking goals of science. Various non-epistemic values play a less clear role. However, non-epistemic values can and should often play a role, for example social, political, or ethical values. But, should these non-epistemic values also play a role in the "heart" of science, f.e. whether to accept or reject particular hypotheses?
Contemporary science can be seen as "mechanistic science" (Barret & Raffensperger, 1999), contrasted with "precautionary science". Precautionary science lessens the frequency of false negative claims, instead of the current science progress with puts more emphasis on minimizing false positive claims. Precautionary science would do so by accepting a wider variety of data, a wider range of harms, and emphasize more complicated and multidisciplinary research projects.
As of now, many of these values permeate science, often implicitly. Lacey (1999) claims that contemporary science is far from neutral with respect to different value systems, because it focuses on employing "materialist strategies".
There are four aspects of nanotoxicology research which Elliott considers, in each looking at the value judgements. These four are:
- materials studied;
- biological models;
- effects examined, and
- standards of evidence
In order to determine what materials to study, many value-laden questions arise. These could be the difficulty of obtaining the materials, the social relevance of investigating those materials, et cetera. These decisions made are not made solely from epistemic values, but incorporate non-epistemic values as well.
One example of a choice of a biological model is what species of fish to use, when attempting to determine the toxicity of particles. Choices must be made between species which make the study more ecologically useful (a prevalent fish), or between fish which are scientifically better understood.
Other aspects are the sensitivity of organisms, which make the results more or less representative.
Examining observed effects, and in what way, also includes making value-laden decisions. For example, determining long-term effects by short-term studies. Or, studying the effect of a pollutant by looking at few organisms in an ecosystem, and attempting to draw conclusions about that entire ecosystem. Choosing to study a large variety of organisms is a value-laden decision.
Often, trade-offs must be made between examining different effects, but the decisions are based on the monetary costs of that study. This is of course also a non-epistemic value.
When do you believe a specific particle is toxic or not? Do you wait until there is enough evidence that it is not, before you say that it is not? Or do you take a "guilty until proven innocent" stance, assuming that it is toxic and only allowing its use until it is proven to be harmless?
Scientists engaged in policy-relevant research cannot entirely avoid ethical and societal value judgements about what standards of evidence to demand - and therefore how to interpret results. It is true that sometimes, scientists can provide uninterpreted data to policy makers. In many other cases, however, it is impractical for scientists to avoid these value-laden decisions.
In some cases, it is unclear to see the social ramifications of making one over the other. In other cases, however, the effects are (or can be) clear. In these cases the question arises: should we intentionally allow ethical and societal values to play a role in these decisions, and if so, how?
Elliott argues that scientists should not ignore the societal consequences of choosing standards of evidence (or any other of the four judgements shown above).
This can be argued for by appealing to social responsibilities that scientists have to society by virtue of their professional role. Alternatively, one can argue for this by showing that scientists are moral agents with the same moral responsibilities as everyone else to avoid negligently causing harm to those around them.
The question how to incorporate these values is a difficult one. Elliott introduces three proposals:
- Providing strategic training in research ethics for scientists working in policy-relevant fields;
- Carefully diagnosing appropriate mechanisms for deliberation between scientists and stakeholders, and
- Supplying significant government funding and leadership for research in nanotoxicology.
In the case of nanotoxicology, research face value-laden decisions about what materials to study, what biological models to use, which effects to examine, and what standards of evidence to demand. In order to make ethical and societal reflection on these decisions more effective, Elliott proposes three suggestions: providing research-ethics training, doing something with stakeholders, and investing in independently funded research.
Diekmann and Peterson look at epistemic and non-epistemic values in (engineering) models. In fact, they make the claim that non-epistemic values are, and should, included in engineering models.
The paper shows multiple examples of where non-epistemic values have influenced scientific or engineering models.
This example concerns the reference dose (RfD) used by toxicologists in risk assessment of chemicals. The non-epistemic goal of the RfD is to identify a dose of a toxic substance that is sufficiently unlikely to put humans at risk. In calculating the RfD, multiple parameters are used which are deemed appropriate because safety is kept in mind during the creation of the model. Safety, however, is no epistemic value. Therefore, many models have parameters which are influenced by non-epistemic values.
This example is about fall detectors used for elderly people. The device detects a fall by recording four phases. If these four phases are detected after each other, a fall is determined. Every phase, however, is detected if the device observes something, and this is compared with some threshold. If it exceeds this threshold, the device 'records' the phase.
The optimal value of these thresholds, however, are no purely epistemic value. In fact, a very sensitive detective would cause a lot of irritation for the patient, and a device which isn't sensitive enough could fail to detect a fall and affect the patient's wellbeing negatively.
Here, wellbeing is a non-epistemic value that sometimes influence the construction of a model.
For the US congressional seats and federal funds, among others, the US population is counted every 10 years. A problem in the census is how to deal with ethnic minorities, which are less likely to be counted than others. The Bureau of Census proposed a model that would lead to statistical improvements, according to them. However, another model was created by a group of independent statisticians, proposed in order to reduce the effect of the undercount problem. A few years before the census, the model of the Bureau was chosen. According to Fienberg (1994), this was not chosen with regard to epistemic considerations, but was entirely motivated by political and economic reasons. If this is correct, then sometimes, the choice of model is directly affected by non-epistemic - moral and political - values.
Now that we know that non-epistemic values often play a role in scientific and engineering models, the question remains whether they should play a role in these models.
Some models have a purely epistemic goal, but sometimes models have, at least partly, a non-epistemic goal. According to Diekmann and Peterson, in these cases, the models should not just focus on the epistemic values, but also on non-epistemic values.
The general argument of Diekmann and Peterson goes as follows:
- Models are, and ought to be, developed with one or several goals in mind.
- Sometimes one of the goals is, and ought to be, a non-epistemic goal.
- The extent to which a non-epistemic goal is accurately reflected in a model depends on the influence of non-epistemic values.
- Some models are influenced, and ought to be influenced, by non-epistemic values.
The view that models should be entirely value-free, or that they should at least contain only epistemic values, fails to recognize that the goals that the models should fulfill are not purely epistemic, at least oftentimes.
In fact, non-epistemic values are in many cases as important, or perhaps more important, than epistemic values. Thus, for models attempting to reach a non-epistemic goal, non-epistemic values should be taken into account.
Longino and Doell look at masculine bias prevalent in scientific research. Hubbard and Lowe show sexist science both as bad science (asking "scientifically meaningless" questions), and as science as usual (generally meaningless). Longino and Doell look at this and see that if it is bad science because of poor methodology, then there is a better methodology to be found that will not reach these biased conclusions. However, if sexist science is science as usual, then we will inevitably reach these false conclusions, unless we change paradigms. By getting a better understanding of the operation of male bias in science, we can move beyond these two perspectives in order to find a solution.
In history, humans have preferred theories or explanations that portrayed themselves in favorable light, similar to the tendency to resist alternative theories. In sixteenth-century Europe, when the geocentric model was 'central' to common physics, there was a lot of individual resistance against the new astronomy (heliocentrism). Similarly, in the nineteenth century, there was resistance to the Darwinian evolutionary theory, as human uniqueness was threatened.
Today, theories attribute the centrality of male development to the development of the human species, or the theories attributing the differential distribution of social-behavioral traits to the differential distribution of certain hormones between the sexes, has much to do with the androcentric and patriarchal belief of our culture. Feminists can identify the influence of patriarchal culture. However, understanding that the theories incorporate a male bias, and understanding how this can be so, are two distinct things.
Facts are an important part in our lives. However, the description of these facts is limited by our sense organs, as well as the language we use to express them. What counts as facts will vary according to our culture and environment. This process of selection the first point of vulnerability to external influences.
Since the same facts can be described differently, there is a possibility for multiple descriptions of a single reality, so a given presentation of data might reflect social or cultural biases. This is another point of vulnerability to external factors.
The category "facts" and the category "evidence" are quite different. The facts we actually know is a function of our perceptual and intellectual structures. Evidence, on the other hand, is facts taken in relation to something else -beliefs, hypotheses, theories. To say that a fact (F) is evidence for a hypothesis (H) is to take F as a sign of H, or to claim that F is so because H is true.
Two studies are looked at: evolutionary and endocrinological studies. For both areas, Longino and Doell attempted to systematically assemble and analyze the material (questions asked, hypotheses found, evidence found for hypotheses, and distance of evidence to hypotheses), to see some of the variety in the ways masculine bias functions in science.
The term "distance" is used to convey the notion of being more or less consequential. The less a description of fact is a direct consequence of the hypothesis for which it is taken to be evidence, the more distant that hypothesis is from its evidence.
Two main categories are looked at: anatomical and social evolution. Anatomical evolution looks at biological changes between humans and other primates, and social evolution concerns interactions among these developing creatures
In recent years, scientific reconstruction of human descent have centered around two focal images: man the hunter and woman the gatherer.
One example of a difference in the androcentric and gynecentric viewpoints is the emergence of tool use. The androcentric view attributes the development of tool use to male hunting behavior. The gynecentric perspective attributes the development of tools to female gathering behavior.
In short, both theories have a similar distance between the evidence and the hypotheses: a lot. Longino and Doell suspect that a less gender-biased theory will eventually supersede both currently contending accounts. The issue, however, is that there is not enough evidence for either of the frameworks. Thus, no framework can be really chosen over the other.
There are three general categories regarding questions about the relation of sex hormones to sexual differentiation:
- Effects on anatomy and physiology;
- Effects on temperament and behavior, and
- Effects on cognition.
This paper looks at the effects on research in anatomy and physiology and in behavior.
Commonly accepted stereotypes of sex-linked behaviors and their presumed correlation with different hormonal levels often provide the starting point and underlying context for more serious scientific explorations, despite the fact that the character of stereotypes preclude their acceptance of genuine data.
The relationship between hormonal changes and sex organ development are clear and well understood. However, issues exist regarding the pathways testosterone follows in producing its physiological effects. The relation between data and hypotheses becomes much more complex in attempts to link hormonal levels with behavior.
This part is relatively less good than the others, relying on the mistakes of Ehrhardt's study (such as extrapolating from rodents), rather than general flaws of the evidence for existing hypotheses. Nonetheless, Longino and Doell claim that there is considerable distance between evidence and hypotheses regarding the hormonal determination of behavioral sex differences, showing that it contrasts sharply with the close fit between the two in the case of anatomical sexual differentiation.
Longino and Doell claim that, since the evidence fails to bridge the gap between the data and hypotheses, the choice of a physiological or an environmental explanation for behavior is subject to the preconceived ideas and values of the researcher.
Longino and Doell look at researchers attempting to demonstrate the observable sexual division of labor has deep roots in the evolution of the species. Evolutionary studies with provide the universals - gender and sex roles have remained constant - while neuroendocrinology provided the biological determination. Longino and Doell show that neither claim need be accepted.
Male bias is present in the evolutionary framework: chipped stones are seen as unequivocal evidence of male hunting only in a framework that sees male behavior as central (to the evolution of the species, and the survival of any group). Neuroendocrinological studies contain sexist bias by linking physiological explanations with androcentric evolutionary account. This possibility has raised concern that some will see a biologically determined human, which includes behavioral sex differences sufficient justification for maintaining social and legal inequalities between the sexes.
Sexism does not seem intrinsic to the interpretation of data as evidence for physiological causal hypotheses. However, sexism is prevalent in the evolutionary science, with the man-the-hunter view being accepted more often than it reasonably can. In endocrinological science, the search for the linking of physiological differences with this prevalent evolutionary view is also indicative of male bias.
Distance between data and descriptive language on one hand, and between data described and hypotheses on the other, leave room for androcentrism and sexism to operate. In evolutionary studies, androcentrism is at work determining the explanatory hypotheses. In hormonal studies, androcentrism is also a possible motive behind the preference for the system of interpretation.
One response is to adopt assumptions that are deliberately gynecentric or unbiased with respect to gender, and to see what happens. Replacing androcentrism with gynecentrism may not be a final strategy, but it does reveal the epistemologically arbitrary nature of androcentric assumptions.
When the issue concerns unreliable but not explicitly androcentric or sexist assumptions, but are suspected of being sexist in motivation, it is important to expose their unreliability and search for additional determinants. Hereditarianism and various forms of biological determinism has been at the service of race and class supremacy, as well as male domination.