Reasoning and diagnostics Part II

In this issue, Barney Donohew explores critical thinking

Published:  13 March, 2017

We began this journey last issue, so to recap: We need solid reasoning skills to carry out effective diagnostics; persistently good decision making doesn't happen by chance. Possibly out of convenience these skills are often underestimated and undervalued by people, both in and out of the trade. We must raise awareness of the discipline and precision of thought necessary for logical and critical thinking: so we can be better rewarded for our efforts; and to make sure they are consistently and properly applied.

Reasoning, arguments and hypotheses
We covered some fundamentals in my last article: we explain our reasoning using arguments, which contain statements supporting a conclusion; one type of argument, a deductive argument, should guarantee the truth of its conclusion (if it is sound); however, we need to use critical-thinking to check this, by making sure i) there are no other possible conclusions (which makes it a valid argument) and ii) the supporting statements are true.

Within our workshops we constantly face plausible but invalid deductive arguments where the initial statements are true but the conclusion is not guaranteed:

1) The vehicle has faults. Vehicle faults cause DTCs. So, the vehicle has DTCs
2) The vehicle has no DTCs. Vehicle faults cause DTCs. So, the vehicle has no faults

Most of us are wise to the unsound conclusions in these examples, although we have not always been and certainly many of our customers are not.

In diagnostics, deductive arguments are best used when we need to totally justify a conclusion after evaluating a specific system or component test result (e.g. A volt drop above 250mV between these two points on this wire indicates a high resistance fault. The Volt drop is 3 Volts. Therefore, the wire has a high resistance fault between these points.)

How do we justify which test to perform next? We do that by first predicting (making a hypothesis) where the fault might lie by using the observations made within our case (e.g. the symptoms, DTCs and test conclusions). We do not know for sure what the fault might be for any given set of observations (otherwise we would not need diagnostic testing!), so we must make an educated guess as to the probable or possible causes. In these cases, we must use inductive or abductive arguments.

Imagine a vehicle reporting a P0671 DTC (Cylinder no. 1 glow plug circuit) within the engine ECU. There are no other discernible symptoms. Where should we start our diagnostic process and how would we justify it?

Very likely we would decide to test the glow-plug supply voltage and current draw, as our experience tells us that when there is only one glow plug DTC then it is likely that the glow plug itself has gone open circuit. This is an inductive argument; written formally it becomes:

Almost every time I have read just one glow-plug DTC from an engine ECU (specifically either P0671, P0672, P0673 or P0674), the glow plug within the indicated cylinder has been open circuit. This engine ECU is reporting a P0671 DTC. Therefore, the glow plug in cylinder no. 1 is open circuit.

From our prior knowledge, we have extrapolated a conclusion predicting the probable cause of an observed DTC; i.e. we have hypothesised a likely fault within a candidate component. From a perspective of critical-thinking we must bear in mind that the conclusion is not guaranteed to be true. This is a feature of all inductive arguments and we can only assess them relative to the strength of their arguments. Diagnostically, we might use our critical-thinking to weigh-up whether the strength of an inductive argument (amongst other factors) justifies us carrying out a system or component test, but it is unwise to rely on it to entirely accept or reject any hypothesis about what is the fault. The glow-plug example could be classed as a mildly strong argument. A stronger inductive argument might be:

Every time I disconnect a CKP sensor on a running vehicle the engine stops. Therefore, if I disconnect the CKP on this vehicle the engine will stop. [I have yet to come across a vehicle where this has not been the case but there remains the (admittedly very slim) possibility that an engine on some vehicle somewhere in the world at some point in time might continue running with its crank position sensor disconnected].

Essentially the inductive arguments above describe case-based reasoning: "This set of symptoms is like those I observed in a previous case. By analogy, the fault might be caused by..." However, experience can be a double-edged sword: we must be flexible in our reasoning and able to revise our beliefs if necessary; especially in the face of rapidly evolving automotive technology, where any given symptom might be caused by an increasing myriad of potential faults.

Even the most knowledgeable amongst us face situations in which they have only limited experience, understanding or exposure; whether that pertains to an entire system, subsystem or the outset of a case displaying a sparse set of symptoms not previously encountered. It happens to all of us. Regularly. So, if, for example, we find ourselves with a vehicle exhibiting some odd, never seen before, behaviour, what is it we do (after checking all the basics, of course) that gets us to the point of selecting a given system or component test?

Well, we might use parts darts or a brute-force attack (sequentially testing each component one by one) or perhaps something better?! How about if we use theories (hypotheses) that relate component faults to possible symptoms and find the best match of these to the set of observed symptoms to try to predict the likely cause of the problem? If this happens to be what you do, then you are doing the right thing; it is a creative hypothesis forming process known as abductive reasoning and is the most skilful and challenging part of diagnostics. Although it is often summarised as "forming your best guess" it is the only diagnostic strategy that, with practice and knowledge, will allow you to complete the diagnostic process efficiently and effectively.

The hypotheses we form must fit within all existing knowledge regarding system and component behaviour and explain the causes of any possible sparse and/or diverse set of observed symptoms, as well as any new observation added to that set (otherwise a new hypothesis will be required).

Think back to the first time you encountered a P0171 DTC (system too lean), still a (infamously) problematic scenario for many. It is the perfect scenario for an abductive approach: for this symptom, we must consider a set of possible faults and their side-effects, e.g. what faults might cause the engine to sense more air than it is expecting (i.e. air intake leak downstream of air-mass meter; under-reading air mass-meter; faulty oxygen sensor; faulty throttle position actuator or sensor etc.) and relate them to the observed DTC and any other symptoms (e.g. erratic engine idle). Each possible candidate system or component will have its own chance of being the actual fault and we must proceed with our diagnostic testing and systematically prove (using deductive evaluation of system and component test results) where it lies.

Messing with memes
Over the last couple of years, a meme has entered the consciousness of many technicians and readers of Aftermarket Magazine: #testnotguess. Whilst appropriate, with our new knowledge, we see that it could be replaced with an alternative meme to more accurately convey the entire diagnostic process and provide positive instruction: #guessthentest. Given that we must use our imagination to create a hypothesis (which we must test), our imagination is a valuable tool within our diagnostic arsenal (and must be recognised as such); if we view guessing as either an act of induction from our prior knowledge or, the more creative, process of abduction, then the meme neatly encapsulates fault-finding.

Summing up
We can use inductive arguments to justify, with some probability, which potential fault might cause any observed set of symptoms based on our past experiences or knowledge.

We use abductive reasoning when we need to predict a set of symptoms, from their hypothesised relationships to possible component faults, that best match (or cover) an observed set of symptoms. Neither reasoning method guarantees the truth of their conclusions and must be validated.

Related Articles


Click here to submit an event


©DFA Media 1999-2021
Terms and Conditions