Tuesday, 10 December 2013

A Bias Smorgasbord

After 3 1/2 years of studying cognitive bias in dogs, this has become a deep and abiding passion. Biases are fascinating and mind-bogglingly prevalent. They help us in many ways, but sometimes they are profoundly unhelpful. This article discusses just some of the many cognitive biases that have been documented in humans. If we know about some of the main offenders, we can decide for ourselves whether to let our biases help us or reject them as a hindrance.

A healthy dose of skepticism is one of the best things you can cultivate when it comes to caring for and training your animals. The reason why is because it is astonishing how what we come to believe can be twisted and warped by bias. And most of the time we have no idea it's happening. The training world is full of ideas that were formed through bias and then supported by bias until they come to represent fact purely on the basis that a lot of people have said it, and some of them sounded like they knew what they were talking about, or were perhaps professionals.

Firstly, it's important to understand that the decisions we make are based on risk assessment. All the biases described below are most likely to occur when the consequences of being wrong are less risky than the consequences of being right. There is a great paper by Haselton and Nettle (2006) that explains this concept in terms of natural selection.
Brace yourselves for a reality check, folks. Generally people have inflated views of themselves, being quite optimistic and considering themselves more in control of events than they are likely to be (Taylor and Brown, 1988; Roese and Olson 2007). Our first question of ourselves in training should be are we as good at it as we think we are? Will we truly see the details we imagine we would see? Is the change in behaviour we think we see real and is it a result of our training? As we explore some of the biases we are subject to, the answer to that last question may become more and more uncertain.

Where events are random to some degree, people believe they have more control over the flow of events than they do in reality, creating an illusion of control (Alloy and Abramson 1979). We develop rituals and a strong belief in our ability to control events (Matute, 1994, 1995). This is where superstitions come in, and, incidentally, obsessive/compulsive behaviours. Animal behaviour is not random, but it is extremely complex and there can be an element of uncontrollability. We should be very careful how much of an animal's behaviour we attribute to our actions. Remember that when we are training an animal, our behaviour is still subject to reinforcement. If the animal does what we intended it to do, it's part of our nature to attribute that to something we did and so we do it again. And once we think we are creating an effect, we will do just about anything to find support for our continued belief in that effect. This is called confirmation bias, where we seek or interpret evidence in ways that support  existing beliefs, expectations, or a hypothesis in hand (Nickerson 1998). For example, success with a training method once may lead to an inflated belief of its effectiveness. You will see success where it doesn't really exist and fail to see problems with the method when they do exist. One obvious reason for this is the desire to find a hypothesis correct or incorrect, usually because people like being right. We'll do just about anything to make sure we are right! Not only do we look specifically for evidence to support us, but we tend to ignore or even avoid evidence that goes against an already held belief (Koriat et al. 1980). Confirmation bias may include selective testing of a hypothesis that will fail to test both sides of the story (Wason, 1960), paying more attention to positive evidence than negative evidence (e.g. Gilovich 1983), and seeing what you are looking for or expect (e.g. Foster et al. 1976). People often draw conclusions early on in a process and then seek evidence to support those conclusions, which is known as the primacy effect (Anderson 1965). What's more, once a belief is formed, it tends to be very resistant to change even in the face of strong evidence against it (e.g. Ross and Leper 1975), and we are entirely likely to take information unrelated to the hypothesis at all as evidence for our continued belief (Pitz 1969)!

On top of this, we have other odd little biases to contend with. We tend to think we are less biased than everyone else (Bias Blind Spot) (e.g. Pronin et al. 2002), which is just gloriously ironic. And we tend to prefer immediate pay-offs over long-term pay-offs (Hyperbolic Discounting) (Green and Myerson 1996 for review) which reminds me of that sublime split second in which a dog pauses in its barking after you shout at it. There's the Mere Exposure Effect, where people prefer things they are familiar with, particularly at a sub-conscious level (Bornstein and D'Agostino 1992), and Negativity Bias, where people pay more attention to negative than positive experiences (e.g. Rozin 2001). And lastly, perhaps the most insidious of biases, Inattentional Blindness, where we just don't see what we're not looking for (see Simons 2004 for a decent review). There are some good videos on YouTube that will demonstrate this. It's a lot of fun and very enlightening, so look them up. Do a search for "inattentional blindness" or "change blindness". The most well known example is embedded below. The things we think we will notice we often don't. That goes for small details and major changes. This could hurt both our training effectiveness and our assessments of the welfare of animals we are training.

These human biases are NOT defects, they are adaptations. We all do it, even scientists. Scientists have just had it drilled into them to look for bias and always try to find something more concrete than their own interpretation of events. Remember that biases have all run through the natural selection gauntlet and ultimately benefit us in many circumstances. They are often like cognitive shortcuts that help us use our experiences to find patterns and make decisions where we don't have a lot of information. But they can be real pitfalls in training and interpreting behaviour. Our animals are at the mercy of our decisions and interpretations, and the way we make decisions and interpret events is coloured by our biases. We can easily be led astray by our natural tendencies to be biased, and our animals may suffer for it.
When trying to prove something scientifically, the general rule is numbers don't lie (although interpreting numbers puts us squarely back in bias wonderland). Our animals can't tell us directly when we are wrong about them, but they can often tell us indirectly. Count the occurrences of behaviour within a set timeframe to determine whether it is increasing or decreasing or staying the same. Calculate success rate. There is a photo of a simple success rate counter on the Creature Teacher Facebook page. Get a stopwatch. Record unwanted behaviour in a diary. Familiarise yourself with an ethogram for your species or make one for your animal. Look for the behaviours in your ethogram, even when you don't expect to see them. That's what we do to convince ourselves in science. It's actually not that hard on a small scale for a layperson, and my guess is it will be a real eye opener. Start with one behaviour at a time and do lots of watching. I think it's a bit more kind to yourself than becoming your own biggest skeptic, which I guess is the second best thing you can do to keep yourself honest.

Alloy, L. B., & Abramson, L. Y. 1979. Judgment of contingency in depressed and non-depressed subjects: Sadder but wiser? Journal of Experimental Psychology: General, 108, 443-479.
Anderson, N.H. 1965. Primacy effects in personality impression formation using a generalized order effect paradigm. Journal of Personality and Social Psychology, 2(1): 1-9.
Bornstein, R. F., D'Agostino, P.R. 1992. Stimulus recognition and the mere exposure effect. Journal of Personality and Social Psychology, Vol 63(4).
Foster, G., Schmidt, C, & Sabatino, D. 1976. Teacher expectancies and the label "learning disabilities." Journal of Learning Disabilities, 9, 111-114.
Gilovich, T. 1983. Biased evaluation and persistence in gambling. Journal of Personality and Social Psychology, 44, 1110-1126.
Green, L and Myerson, R. 1996. Exponential Versus Hyperbolic Discounting of Delayed Outcomes: Risk and Waiting Time. American Zoologist 36(4): 496-505.
Haselton, M.G., Nettle, D. 2006. The paranoid optimist: An integrative evolutionary model of cognitive biases. Personality and Social Psychology Review 10(1):47.
Koriat, A., Lichtenstein, S., & Fischhoff, B. 1980. Reasons for confidence. Journal of Experimental Psychology: Human Learning and Memory, 6, 107-118.
Matute, H. 1995. Human reactions to unavoidable outcomes: Further evidence for superstitions rather than helplessness. Quarterly Journal of Experimental Psychology, 48, 142-157.
Nickerson, R. 1998 Confirmation Bias: A Ubiquitous Phenomenon in Many Guises. Review of General Psychology 2(2):175-220.
Pitz, G. F. 1969. An inertia effect (resistance to change) in the revision of opinion. Canadian Journal of Psychology, 23, 24-33.
Pronin, E., Lin, D.Y., Ross, L. 2002. The Bias Blind Spot: Perceptions of Bias in Self Versus Others.Personality and Social Psychology Bulletin 28(3):369-381.
Ross, L, Lepper, M.R., and Hubbard, M. 1975. Perseverance in Self-Perception and Social Perception: Biased Attributional Processes in the Debriefing Paradigm Journal of Personality and Social Psychology 1975, Vol. 32, No. 5, 880-802
Roese, N.J., Olson, J.M. 2007. Better, Stronger, Faster: Self-Serving Judgment, Affect Regulation, and the Optimal Vigilance Hypothesis. Perspectives on Psychological Science 2 (2): 124-141
Rozin, P. 2001. Negative bias, negativity dominance, and contagion. Personality and Social Psychology Bulletin (5)4: 296-320.
Simons, D. 2000. Attentional capture and inattentional blindness. Trends in Cognitive Sciences 4(4): 147-155.
Taylor, S. E., & Brown, J. D. 1988. Illusion and well-being: A social psychological perspective on mental health. Psychological Bulletin, 103, 193-201.
Wason, P. C. 1960. On the failure to eliminate hypotheses in a conceptual task. Quarterly Journal of Experimental Psychology, 14, 246-249.

No comments:

Post a Comment