3.6 General discussion
3.6.5 Final thoughts and implications
Why, ultimately, has the research community been so interested in selective exposure? I think it is because it is indicative of a broader tendency that seems problematic: the
tendency to reason in ways that lead us to confirm what we already believe, and that prevent us from changing our minds even when doing so might be important. As I
said at the beginning, selective exposure is just one form of a broader tendency known as confirmation bias. In more intuitive terms, selective exposure is interesting to us
because it seems to indicate a lack of open-mindedness, and a lack of open-mindedness seems like something to be concerned about. In turn, the reason we are concerned
about confirmation bias and/or closed-mindedness is not purely intellectual - they seem to have certain, concerning, practical implications - preventing us from figuring out the
truth on certain important scientific and political issues, fueling conflict, and hindering progress. I will discuss these issues in more detail - particularly concerns around a lack
of ‘open-mindedness’ - in chapter five.
The selective exposure literature, however, as it has got caught up in this question of
whether and to what extent the effect really holds, has begun to lose sight of this wider question. Selective exposure in itself is not particularly interesting - it’s interesting
insofar as it contributes to confirmation bias and/or a lack of open-mindedness. But as discussed above, as it is currently studied, selective exposure isn’t actually a very
good measure of bias or closed-mindedness: it’s possible for an entirely rational person to exhibit selective exposure under certain assumptions, and someone might seek out
counter-evidence but still approach it with an entirely closed-minded attitude. Selective exposure also hasn’t been studied in a way that easily allows us to draw conclusions about
its practical implications - that is, it’s hard to say from these very abstract lab studies whether a tendency towards selective exposure, if it exists, causes genuine problems for individuals or society.
I think the reason selective exposure effects have been so mixed and elusive is that there are many different factors influencing someone’s choice of information: their motiva-
tions, what they already know and believe, their expectations about different sources of information, which are difficult if not impossible to capture experimentally. The reason
that mixed effects for selective exposure have been viewed as so surprising is that it’s been assumed that selective exposure broadly corresponds to ‘open-mindedness’ (and
it’s broadly assumed that people are not open-minded.) This stems from a failure to recognise that without understanding these motivations, beliefs, and expectations, it’s
very difficult to draw conclusions about open-mindedness or rationality from findings of selective exposure (or a lack thereof.)
In studying selective exposure, I suggest, we need to take a step back: and think about these findings with a clearer understanding of other reasoning processes, and more back-
ground of what it means to be biased, what it means to be open-minded. Rather than asking “is there evidence for selective exposure?”, we instead need to ask questions like,
“what does it really mean to be open-minded, why is this important, and how do we measure it?”, and “to what extent is confirmation a problem, and how much of this
comes down to how people select information versus bias in other stages of the reason- ing process?” I’ll now turn to these broader questions: first looking in more detail at
what it really means to be biased or irrational, and the implications for confirmation bias, before turning to a closer look at the related concept of open-mindedness.
Bias, rationality and improving
human reasoning
4.1
Introduction
I’ve reviewed a variety of research claiming that a ‘confirmation bias’ arises in human
reasoning - and discussed some reasons that the evidence for confirmation bias is less strong than it first seems. A theme we keep returning to is the difficulty of establish-
ing standards by which to evaluate how people should reason - and how most of the research on confirmation bias fails to adequately address this. Ambiguous use of terms
like ‘biased’ or ‘rational’ also seem to create unnecessary disagreement and confusion. This chapter therefore focuses in on these normative issues, clarifying some of the ter- minological confusion to see what substantive disagreements we are left with.
Many others have acknowledged the difficulty of establishing when and how it’s rational
to reason in ways that ‘confirm’ what one already believes. Nickerson says at the end of his review on confirmation bias that “the question of the conditions under which one
should retain, reject or modify an existing belief is a controversial one”, and that “it is natural to be biased in favour of one’s established beliefs... whether it is rational is a complicated issue that can too easily be treated simplistically.” (Nickerson, 1998, p.209)
In many classic studies of confirmation bias, the normative standards against which performance is compared are vague, disputed or nonexistent, and several arguments
have been made that tendencies classically interpreted as evidence of confirmation bias
may actually show no such systematic bias when these normative standards are made clearer (Austerweil and Griffiths, 2008, Jern et al., 2014, Klayman, 1995, Klayman and
Ha, 1987, Oaksford and Chater, 1994, Perfors and Navarro, 2009). Klayman concludes that “it is quite clear that quite a few of the putative sources of confirmation bias
do not directly imply any consistent bias towards the focal hypothesis.” (Klayman, 1995, p.398) Even if a tendency to favour what one currently believes does exist, others
still have argued this might be rational if we assume people have different goals than simply epistemic truth-seeking (Friedrich, 1993, Mercier and Sperber, 2011, Tooby and
Cosmides, 1992).
Though it’s widely acknowledged that these complex normative issues exist, it’s still
not clear what the implications for confirmation bias are. Many authors continue to talk past each other, using terms like ‘biased’ and ‘irrational’ to simply mean different
things. Nowhere in the literature are the different positions on what these terms mean laid out clearly so that we can see how disagreements might arise in different places,
and what their implications are. My aim in this section is therefore to lay out as clearly as possible the different ways in which people might disagree about the ‘rationality’ of
confirmation bias.
Initially, this will involve taking a step back and looking at how normative issues have
been discussed and debated in the psychological literature more broadly, though I will return to examples in the confirmation bias literature throughout. At the end of the
section, I will look more specifically to the question of how this impacts confirmation bias.
Of course, getting clearer on what we mean by terms like ‘bias’, ‘rationality’, and how people ‘should’ reason is important not just for understanding confirmation bias, but
for psychology as a whole. A great deal of the psychological research conducted over the past 50 years has painted a picture of human reasoning as prone to bias and error,
of human beings as irrational. The natural response to this is to ask: how might we fix these biases, how might weimprove human reasoning? Particularly as it begins to look like aspects of human irrationality may contribute to real-world problems, the project of improving human reasoning and reducing bias is potentially an incredibly important
one (Larrick, 2004, Lilienfeld et al., 2009). However, it is also a much more complex task than it first seems - there still isn’t really a clear consensus on what it even means
to be biased or irrational, let alone an answer to whether it is possible (or desirable) to improve things.
This chapter will be structured as follows. First, I will review some disagreements in the psychological literature about how to apply different terms and normative standards,
and how these disagreements affect the confirmation bias literature. In particular, I will discuss three things: how the term ‘bias’ has been used in different areas of research;
how the normative models that provide the standard against which bias is measured are justified; and further confusion arising from different notions of ‘rationality’. I will
attempt to clarify where there are really substantive disagreements about confirmation bias and rationality, as opposed to mere terminological confusion. I will then ask why all
of this matters - arguing that these subtleties in what it means to be biased or rational are not mere pedantry, but have important implications for how we think about the
possibility of improving human reasoning. Finally, I will summarise the implications for the literature on confirmation bias more broadly.