Training Gubun, my dog and the auto-drivers of Assam
How behavioural ecology and economics help me train my dog and errant auto-drivers
Learned behaviours play a critical role in the survival of animals by securing access to food, shelter, and safety. Anyone who has had a dog will know how quickly they pick up on human cues and even predict our behaviour – including things that are not hard-wired. For instance, the crinkling of wrappers will prick up the ears of dogs (and little children) -- clearly, something that evolution has not explicitly selected for - unlike a spider spinning a web which has a strong genetic basis.
As Tejaswini, Rangjalu, and I rescued and brought up the tick-and-flea-ridden stray puppy from the banks of the Brahmaputra river, we got a fantastic opportunity to train her. We named her Gubun, meaning second in Bodo.
Gubun right after we rescued her
Obviously, dogs do not understand the literal meaning of words, but they very quickly form associations between sounds and related actions. Training her was an absolute delight particularly using principles from behavioural ecology and behavioural economics – two fields that I have been dabbling in as part of our work with human-elephant conflict.
Human-elephant conflict is a serious issue in northeast India and needs a good understanding of human behaviour
Training her to ‘sit’, ‘stay’, and ‘come’ was the first set of commands. Now to train her, I needed to make sure that the association between the words and the action was reinforced. Therefore, I would get her to sit, giving the command ‘sit’ and following it up with a treat (she loves fatty catfish). Once this link has been established, I can get her to sit just by saying ‘sit’.
This is identical to one herd of elephants moving away when they hear the sound of a Bolero car that the Forest Department uses when they are chasing elephants away from human areas. Since we use the identical make of car, a Bolero, the elephants move off on hearing our vehicle approaching as well – probably because they associate the sound of a Bolero with noise, fire-crackers, and the general commotion that the Forest Department uses to chase them away.
The Bolero car that the Forest Department uses to get around when chasing elephants away from human areas. © Mahindra
Similarly, Gubun quickly learnt that ‘sit’ meant a treat but only if she was sitting. This meant that I had to resist puppy eyes and not give her a treat if she defaulted and got up. Now, this is a critical step – else she would not learn as then she would associate a treat with the sound ‘sit’ irrespective of what she was doing. These kinds of feedback loops are critical for the learning process.
A team of autowallas taking a break
Therefore, now when I stop an auto – I ask them the price (as is customary). The autowalla (auto-driver) generally states a fair price (probably because this is very close to a perfect competition market – where the consumers have complete information about the price, there are other autowallas to choose from, and the service is uniform, ie, all autowallas will provide an identical service).
However, at times, they decide to take a gamble – and quote a higher, inflated price. Their implicit assumption is that the person could be new to the region and hence does not know the established, fair price. And it makes sense – if the person is unaware then the autowallas stand to make a quick buck, and if s/he is familiar – then they will haggle it down. However, this often wastes time and leads to the inconvenience of others – as this happens in a busy marketplace, holding up traffic. Additionally, the notoriety of autowallas and cabbies leads to a loss of reputation of the town, a problem that the society collectively bears.
Dismantling the Auto-man empire: A descriptive model on the two options before me and their implications on the autowallas learning
Therefore, now, when an autowalla asks me for more than the normal rate in an attempt to rip me off, I refuse. When I do this, they invariably reduce their price to the fair price or start negotiating. However, to drive the feedback loop, I refuse, even when they negotiate to the fair price. This pain of having lost business, I feel, would translate to them not attempting the inflated price right at the outset.
If I chose option II, where I would agree at the fair price after initially being quoted the inflated price, then there would be no learning -- the autowallas are better off trying with an inflated price – as the fair price is, in some sense, already secured as the worst-case scenario.
However, to drive the feedback loop, I refuse, even when they negotiate to the fair price.
A large wild tusker moves away on hearing the sound of a Bolero car in a tea estate
Now the question is, how many of these iterations are required for the autowallas to learn? That is difficult to tell and depends on a host of factors like culture, individual personalities etc. However, Gubun, and the elephants learning to navigate and adapt to human behaviours could provide some clues. For instance, we certainly know that this process does not need to take place all the time to learn, in fact, the unpredictability of the system adds robustness.
Because the elephants do not know which Bolero means trouble, they tend to move away on hearing the sound of any Bolero. Because Gubun does not know when is she going to get rewarded for sitting when she hears ‘sit’, she sits anyway. Similarly, because the autowallas do not know which customer is going to reject them for quoting an unfair, inflated price will lead to them being more fair at the outset. This unpredictability and spurning do not stick out like a sore thumb or seem like an outlier – because seeking ‘fairness’ is a well-established human trait.
Training Gubun elephant commands
Of course, this process of deciding a price is more complex and chaotic than what is described here – but these processes certainly influence outcomes. For instance, when talking to Buddhi, an autowalla about his business, he grumbled, "when it is late and I know that there are fewer autowallas around, I try to charge more -- if I do this at other times the customer immediately goes to the next one...".
Training Gubun has been super fun, also because she doesn’t ‘know’ any languages – so she has now been trained to also follow commands that elephant handlers, mahouts use for their elephants!