reasoning for architects
DESCRIPTION
Translating business intent into technology solutions is a complex cluster of cognitive activities expected from an architect. To perform valid and traceable requirements analysis, architects need to be strong in logical conclusions and critical thinking. Without strong reasoning skills, architects often revert to marketectures or leaps of faith where technical recommendations are stated without explicitly traceable logical chain. This session will show best practices in premises examination, facts validation, checking for logical fallacies and packaging/presenting the conclusions or outcomes.TRANSCRIPT
Reasoning for ArchitectsMiha Kralj
Argument Structure
Think of assumptions as the glue that holds the evidence to the conclusion.
Argument: a claim (or conclusion), supported by evidence.
AssumptionEvidenceConclusion = +
What is not an Argument
Don’t waste energy trying to win debate on non-arguments.
Assertion: Separation of concerns is key SOA principle.
Statement: Zachman said, “focus on Why, How, What, Who, Where and When.”
Explanation: Because there was not enough time, the design is less than optimal.
Opinion: TOGAF framework is a waste of time.
Identifying the Conclusion and Evidence
Think of assumptions as the glue that holds the evidence to the conclusion
Conclusion:
As a result…Clearly…Consequently…Hence…In conclusion…Therefore…Thus…So…
Evidence:
As indicated by…As shown by…Because…For…Since…Given that…As…The reason is that…
Evaluating the Argument
Best practice: make assumptions explicit and write them down
Testing the EvidenceTesting the Assumption
As this PoC in the cloud works really well, cloud computing is surely the right strategy for your company!
This is the most innovative cloud platform; there are more apps currently built on this platform than on any other cloud technology.
Reasoning Flaws
Most Common Reasoning Flaws
Flawed reasoning turns architecture into marketecture!
Over-genralizatio
n
PoorAnalogy
SelectivePerception
Wrong Causality
Ignoring the
Evidence
Selective Perception
Architects should not be influenced by stereotypes, prejudices or preconceived notions.
Tendency to see the world the way we would like it to be rather than how it really is.
We came in, saw the root cause of problems and rapidly deployed our standard offering that was the most efficient solution for them. Competition was stunned!
They didn’t listen, had a preconceived view of the problem, deployed a canned solution and thought we were happy. We will never call them back again!
Poor Analogy
Advice: meaning and scope of words and terms in arguments must be consistent.
Comparison of items made on the basis that they share some similarities
Why go with such radical design change? What worked in the past will work in the future!
She did such a great job as a delivery architect that we’re going to promote her to an architect manager!
Representativeness Assumptions
Don’t test if a sample is large enough – test if it is diverse enough.
A sample should be a valid representation of the larger group (population)
We virtualized all web servers with no problems, so we can proceed with virtualization of the whole datacenter. (statistically accurate?)
I highly recommend this IT vendor - on two projects that they were engaged on, they were really good. (anecdotal evidence?)
Selecting “Good” Evidence
Slanting the evidence is the most common reason for broken advisor relationships
Choosing relevant evidence that supports our stance while ignoring refutable ones.
All our case studies show very strong ROI when moving applications to the cloud!(do we publish ones with bad ROI?)
As a technical reviewer, I can say that every project without strong risk mitigation plan fails!(how about projects that didn’t go through technical review?)
Causality Assumptions
Mixing correlation with causation is the most common reasoning flaw in architecture
Incorrect identification of coincidence, correlation or causation of two events
CIO left the company just before our project was cancelled. Therefore, CIO departure caused us to lost the engagement. (coincidence or correlation?)
He is a hard-working architect. No wonder he gets the toughest assignments in his sector! (reverse causation?)
Techniques
Convergent vs. Divergent Thinking
Use convergent thinking for analysis. Use divergent thinking for innovation.
Divergent:
Opens the mindFloodlight thinkingHolistic and diffusedCreative approachInnovative and emotional mindsetSynthesis
Convergent:
Focuses the mindSpotlight thinkingSequence and orderAnalytical approachLogical and scientific mindsetClassification
Thinking-killers and Thinking-growers
Thinking-killers prevent commission errors but encourage omission errors.
Growers:
Are there any questions?What would happen if…Why is this relevant?What have we missed?Wouldn’t it be fun if…What idea have you come up with?Thank you
Killers:
We tried it beforeThat’s not the job of an architectThat’s not how we do itThat sounds crazy to meThis is not importantThat’s a stupid ideaNo
Pros-and-Cons Analysis
Architects often fall into a fallacy of false dilemma when doing Pros-and-Cons analysis
Middle viewContra sidePro side
The Devil’s Advocate Technique
Architects use the devil’s advocate technique to impose objectivity
Approach of advocating an opposing or unpopular cause for the sake of argument or to expose it to a thorough examination
What would an architect from competitor say if they’d see this design proposal?
So, is it true that this solution would work better/faster/cheaper if we’d use competitor’s product?
Prisoner’s Dilemma
The Prisoner’s dilemma is an example when cooperation is superior to competition
Partner cooperates
Partner doesn’t cooperate
I cooperate with partner
I win a little,partner wins a little
I loose,partner wins
I don’t cooperate and do my thing
I win,partner loses
We both lose big time
Mixed-motive game: both parties can do well if they work together by cooperating or can try to gain an advantage by competing
Fallacies
Fallacies Based on Language
Equivocation:Mainframe is better than nothing. Nothing is better than cloud computing. Therefore, mainframe is better than cloud computing.
Distinction without a difference:We didn’t miss the milestone. We just stretched it for a week.
Amphibology:We will waste no time reviewing the reference design.
Fallacies Based on Bad EvidenceHasty Generalization:
I tried to implement SOA three times and each time it was a disaster. SOA is terrible!
Circular Reasoning:Architects do only architecturally-relevant things. Architecturally-relevant things are what architects decide to do.
Negative Proof:No customer complained about knowledge of our consultants, so we are confident that we are offering more than enough technical readiness.
Ad Hominem:How can you let this architect do the design? He is not even certified!
Fallacies Based on Bad Evidence
.
Poisoning the Well:You are not credible in EA – you are from a country/region where nobody knows how to do EA!
Tu quoque:How can you say that I shouldn’t use TOGAF? You are using it all the time!
Red Herring:You have hard time analyzing requirements? When I was your age, we had to use paper and pencil to perform proper analysis!
Ad Antiquitatem (appeal to tradition):We always used this framework since 1998 – there is no need for a new one.
Fallacies Based on Bad EvidenceAd Verecundiam (appeal to authority):
Our Chief architect says he likes it, so it must be good!
Ad Populum (appeal to public opinion):Everybody that I asked agreed that encryption is not required in our case.
Ad Misericordiam (appeal to pity):They will fire me unless you approve this work order!
Ad Metum (appeal to fear):You don’t want to know what will happen if you allow iPads to connect to that environment.
Fallacies Based on Flawed AssumptionsFalse Alternatives:
You either move IT to the cloud or you waste money with your datacenter.
Golden Mean:Agile was too risky, waterfall was too slow so best for us is a hybrid agilefall methodology
Fallacy of composition:Joe is a great consultant. Frank is a great architect. They will make a great team!
Fallacy of division:This team performed really well, therefore everyone in this team was a great performer.
Fallacies Based on Flawed AssumptionsFallacy of the Continuum:
Win 3.1 was good, Win 95 was good, Win XP was good – therefore Vista was good too!
Incorrect attack on generalization:Really, Windows is good OS? My mother can’t do a single thing with it!
Distortion:What you are saying is you don’t care about architecture and only want deployed solution!
Fault Analogy:We are both architects and PMs. You are a good architect, therefore you must be good PM.
Fallacies Found in Deductive LogicAffirming the Consequent (if A, then B -> if B, then A):Whenever I work on an engagement, I write a lot. I write today, therefore I must be on an engagement.
Denying the Antecedent (if A, then B -> if not A, then not B):Whenever that partner is involved, things are difficult. This time partner is not involved, so everything will run smooth as.
Argumentation
Packaging the Response
A fool with a tool is still just a fool.
Using flawed arguments as an advantage
Allow graceful retreat and correction
What would you say if you’d be a
lawyer?
Why Should I care?
Doubt everything.
Critical Thinking
enables bullet-proof
conclusions
Skepticism distinguishes
architects from
marketeers
Reasoning is the strongest tool architect can master
Argumentative Logic
helps you with traceability