Commonsense-Connections

Here, we discuss a “mental algorithm” for figuring out the meaning of terms involving a connected pair of words. E.g. – Washing soda / baking soda, Hyderabad chilly, (romantic) love-birds, arranged marriage, house-arrest etc. 

That is, a procedure for mechanically figuring out the meaning of such words. It is claimed here, to be also similar in many aspects, and in parts to the way the mind works for solving such problems, in early years (i.e. when children come across terms for the first time). And it can also be incorporated in a machine. 

(Note : We are only considering commonsense-connections which young children can make.

Also, it is assumed you know the meaning (one meaning at least) of both the words). 

Consider Baking soda – 

STEPS

  1. Firstly, the word is mainly the second of the 2 words. That is, here – ‘soda’. Baking soda is a soda, it is not a baking (say, a process/style… of baking). 

So the first question is – WHAT soda? This ‘WHAT’ is the general, broad query about soda. That is, what kind of soda, what colour soda, what property-soda etc. All this is summed up in ‘WHAT soda’?

  1. It is Baking Soda.
  1. So there is soda, and there is the process of Baking. The process of Baking has to be connected to soda.
  1. How can soda be integrated in Baking? (How can the second word be integrated into the first?)

Point : There is a point worth mentioning at this stage. Logically there are infinite ways of connecting the 2 words. But they should make common sense. It should be something similar to something you have already seen or experienced. For example, if someone asks a child what a bird-bag is, one answer is a bag with a bird attached to it. But this isn’t something that is seen/experienced and existing commonplace. (). 

  1. Here come cases. Firstly, there are 4 cases of the word-pairs. 

i)   Physically Picturizable – Physically Picturizable (Washing soda)

ii)  Physically Picturizable – Abstract (House arrest) 

iii) Abstract – Physically Picturizable (Love birds)

iv) Abstract – Abstract (Arranged marriage)

The first 2 are similar because the first word in them is Physically picturizable whereas the last 2 are similar because the first word in them is Abstract (non-picturizable). You can physically picturize the process of washing and obviously also soda. But you cannot have a physical picture of love (or say, things like day, occasion, weight etc.)

  1. If the term is of types (i) and (ii), then try to literally physically-cum-pictorially insert the second word into the physical picture of the first word, in the sense of actual CONTAINMENT, TOUCHING or SURROUNDING NEARBY. (Insert it into the picture). Try possibilities. 

Check for commonsense – whether such a thing exists. (By the way, commonsense is the whole theme of the entire exercise).

  1. If the term is of types (iii), then

b.1) ANALYSE the first word as to – what all aspects are there to it?

            b.2) Take an aspect

            b.3) Check if the second word has an almost literal/actual role-playing place as it (in

            that aspect, or in the first word by itself). (As mentioned at the intermediate ‘Point’, it should make common sense and not just be only logically valid).

  1. If the term is of type (iv), then 

c.1) ANALYSE both the words as to – what all aspects are there to each? 

c.2) Take all possible pairs of aspects (This is where the second of the 3 parts of the claim mentioned at the very beginning of this document – algorithm for how the mind works – takes a beating).

c.3) Check if the second word has an almost literal/actual role-playing place as it (in

            that aspect, or, in the first word by itself). (As mentioned at the very beginning, it should make common sense and not just be only logically valid).


Lets see examples (from STEP 5 onwards) – 

  1. Baking soda – we continue our example above. Washing soda is of type (i). So try to literally physically-pictorially insert ‘soda’ into the physical picture of ‘the baking process’, in the sense of actual CONTAINMENT, TOUCHING or SURROUNDING NEARBY. (Insert it into the picture). 

Imagine the picture of the process of baking. And imagine soda. Where can the soda fit in? Fuel for the fire? NO. The correct possibility is soda used (as/) in the subject being baked.

  1. Hyderabad chilly – This is again an example of type (i). Picturize Hyderabad and insert the chilly into the city of Hyderabad. So, a chilly from Hyderabad. 

Or being more specific, one may reduce it to – What does the picture of Hyderabad, in general, contain? Roads, fields, Government, people, buildings etc. So chilly from Hyderabad fields (lands).

  1. House arrest – This is an example of type (ii). So imagine an arrest inside (a picture of) a house. So an arrest inside a house.
  1. (Romantic) Love-birds – This is an example of type (iii) since love cannot be physically picturized. So, take aspects of love – 1) 2 agents 2) feelings in their hearts about each other 3) being together etc. The second word – birds – fits role-playing-ly exactly into ‘2 agents’ (aspect 1). So, agents involved in love. (The birds (people) are the agents in love). 

(*Another possible meaning is – 2 (actual) birds in love. The second word – birds – fits role-playing into 2 agents, making it into 2 actual birds in love.)

  1. Arranged Marriage – This is an example of type (iv) since both cannot be physically picturized. 

So take aspects of Arrangement – 1) agents 2) relative positions 3) connections etc. 

And the aspects of marriage – 1) 2 agents 2) they living in a legal staying-together relationship etc. 

The agents who are marrying fit role-playing-ly in the agents of arrangement (aspect 1 of the latter fits into aspect 1 of the former). Hence agents (2 people) arranged into a marriage.

Commonsense-cum-Linguistic constraint dataset

For all Language Technologies-related systems, there needs to be a general dataset of the following kind –

We need to build a commonsense-cum-logical-constraint-set – a dataset wherein we link every word (starting with common nouns) in the dictionary with a general type of a constraint word / phrase which is mostly likely to be around that word in a text in which it occurs i.e. somewhere in the sentence or paragraph or……so on….(or probably definitely in the write-up as a whole) – somewhere fairly around.

For example (common nouns) :

  • Smartest (or any superlative word) – Being superlative, there has to be the phrase ‘in something’ i.e. ‘in some set/pool’, around it.
  • Name – there has to be an entity around whose name is being talked about / mentioned.
  • Fund – there have to be some number and currency unit, around.

(This will have to be done manually for words in the dictionary; the project can be crowdsourced to school students over the web).

How we consume reality commonsensical-ly in everyday life

How we consume reality commonsensical-ly in everyday life – 

Before coming to the figure –

Consider any data. What is usually around it, in typical everyday scenes, can be called the “commonsense around it”. Or, in other words, commonsense makes us quickly assume around a piece of data, what is usually present around it, in typical everyday situations.

Here is the explanation of the figure :

  • The rectangular frame is the reality. The circles (red and green) are the data-points in the reality.
  • The red circles are the few data-points we pay conscious attention to. (We don’t pay attention to the whole bit-by-bit reality. We miss many things).
  • The blue amoebae around the solid circles are the commonsense assumptions about / stretching from, the red circles.
  • The green circles inside the blue are the parts of the reality which match with the commonsense-space around the red circles.
  • The amoeba-ically-bordered blue space is the total reality we effectively perceive, from the attended data-points (red), by stretching things around them commonsensical-ly.
  • The green circles outside the blue amoebae are the reality data points we miss – miss conscious attention towards.

Non-contextual commonsense implications

Non-contextual (not any particular context) commonsense implications – 

In any scenario, there are multiple commonsense implications recognizable – only one or two are “mainstream”, the rest are “less effective”. The mainstream ones carry more default relevance than the less effective ones. What is the distinguishing criteria/property? We shall see.

E.g. – 1) If you give a book to someone, it won’t remain with you.

Implication – it won’t remain with you – this is mainstream.

Less effective ones – you will lose contact with the cover of the book

                                 the distance of the book from your body would most probably increase

                                 ….etc

2) If you put a very heavy weight on the head of a bicycle-rider, will the bicycle be able to bear the weight ?

Mainstream implication – above (will the bicycle be able to bear the weight?)

                                        similar one – the weight on the bicycle will increase

Less effective ones – the weight will touch your head.

3) If you open a book, you will see some text/contents inside it.

Mainstream – above

Less effective – the expanse of the book will increase

                        you will see 2 pages of the book (this is not so bad)

                        you will see the binding/partition line of the book.

                                  ~~~~~~~~~~~~~~~~~~~~~~~~~~

What is the difference between the mainstream ones and the rest?

The former shows the relation involving all (as many) the components of the whole scenario (as possible.)

In example 1, the mainstream implication includes you and the book. The less effectives – i) does not include the whole book (only the cover). ii) talks about your body only, and not (you the identity + body) which is = you.

In example 2, the mainstream implication includes weight (rider + weight on top of his head) + bicycle. The less effective one does not include the bicycle.

In example 3, the mainstream implication includes book + its contents. The less effective ones – i) talks about the outline / area of the book  ii) talks about (only) 2 of the pages iii) talks about a part – the partition line – of the book. 

So there is a tie between any of these 3 said-to-be less effective ones and the mainstream one. But the contents of the book clearly carry more default relevance than any of the other 3 parts – 2 pages, the binding line, and the outline.

                                                   ************************

The Heart of Commonsense Thinking

Here, we are talking about problems involving thinking; not pure knowledge-based answering (like say, me asking you – What is my uncle’s name?)

There would be various possible theories to explain the mechanism of the Cognitive Psychological process involved in coming to the solution of these problems involving thinking. Broadly and vaguely speaking, there would be concepts like sequence of events, sub-problems, objects/agents – doers and receivers, cause-effect, if-then, rules, cases, connections and relations etc.

CLAIM – 

Whatever be the theory, here is, according to me, a central heart-like feature to it – The solution can occur to the mind only if one already knows of a state which is like the imagined state.

If the imagined configuration-state involved in the question isn’t KNOWN through sheer raw memory (beforehand), the process of thinking to the solution is impossible.

How can you think about the future, in general, otherwise? What will happen to this rod on pushing it? How can I answer without having known of a “pushed rod”?

Commonsense thinking is unorganised and messy. Hence it suits it to have something like a key feature/”heart”. 

Illustration – 

Consider a horizontal maker pen. It has the word ‘Marker’ written on it along its length. Question – If I tilt the pen, what will happen to the text (‘Marker’) written on it? Answer – The text will also get tilted.

How does the mind process this?

The point is this – You cannot answer the question without having had seen a tilted object with something written on it. 

If you somewhere in the mind (not consciously aware, upfront) don’t know of the state of a tilted object with something written on it, from your commonsense memory, the solution, whatever be the path of your solution, won’t flash. 

(NOTE : The generalization involved in the writing above (like ‘object’ instead of ‘cylindrical object’, which a marker pen is) pertains to cutting out the irrelevant parts of the problem and preserving only the relevant ones.) 

So, speaking in a loose and light sense, you know all future beforehand! 

Why Commonsense expressed in Logic is a bad idea

Why Commonsense expressed in Logic is a bad idea – 

Suppose I say to you – 

X > X – 5        since anything will be greater than something removed from it. This is commonsense thinking. The basis of the proposition here is commonsense.

Now, one might argue that – 

Logically, 

0 > -5

add X to both sides,

so, X + 0 > X -5

so, X > X – 5

Hence, the same thing can be arrived at and represented logically. 

But logic doesn’t capture the reasoning process / the rationale in the above instance of commonsense thinking – that anything will be greater than something removed from it. This is fundamentally different from the logical process of deduction shown above. It (the former) comes from human experience of reality – all that we have experienced about ‘amounts’ in our lives, till the point of putting forward that propositional basis. But all that can be challenged logically by saying that – all this (experience) is finite and only up-till now; how do you know about other cases and the future ? Also you don’t have a record of all the instances of dealing with amounts in your life; may be some skipped from memory. Still, that commonsense thinking piece remains the strongest perceptual force we have of being sure about what we are doing. Putting another way, just imagine how strange it would be if someone doubts that reasoning and believes only in the logical deduction shown above. That would amount to no sense of real understanding / cognition.

The Synchronization between Linguistic and Cognitive faculties

Consider these questions – 

Which year was world war 2 fought in?

Where is the Eiffel tower?

What is Trump’s brother’s name?

Now, in trying to come up with the answers to these questions, one has to tap ones mind with pointers like – ‘year…year….year…..whats the year?’, ‘place….place place….whats the place?’ and ‘name…name…. name…whats the name?’ respectively. One tap’s ones mind’s memory with these objective, definitive, “tangible” items/pointers (year, place, name) – and these questions listed above are indeed KNOWLEDGE-based questions.

Now consider this question – 

Why is the COVID situation so bad in the US?

Here, one has to tap one’s mind with ‘why ….why….why?’ Now ‘why’ is not an objective, definitive, tangible “item”. (That is, ‘reason’). It doesn’t have the same “item”ish flavour as that of year, place or name. You cannot “statically itemize” a ‘why’ or reason or thinking. And indeed, these are THINKING-based questions.

(Loosely speaking, one can “see” knowledge, but not thought)

The Cognitive Birth of a Question

                       The Cognitive Birth of a Question  

This is an attempt to enumerate the kinds of questions as per the basis of their occurrence in the mind. 

Let us divide them into 3 broad categories – 

  1. Thinking-process-based

These are aroused as thought processes. 

Illustrations of the above 2 types –

i) The first-type above – Suppose you enter my room and see a huge and thick pen on the desk. A question will arise in your mind – Why is this pen so big? Here you have identified the entity (the whole – the pen) and the mind has gotten quizzed about its property, as to – why is this property so?

ii) The second-type above – Suppose you see some things together – say some crystals with a blue liquid between them and some glittering material sparkled over the whole. You would get quizzed – What is this? Here, you have identified the parts (the properties of the whole – the crystals, the blue liquid and the sparkling glitter on top), but haven’t figured out/understood what it is (as a whole).

  1. Knowledge-procurement-based

These are primarily about knowledge-procurement. 

These can also be simply called as curiosity-questions. 

Illustrations of the above 2 types – 

i) The left-hand-side type – What is the capital of Serbia? This is purely a “knowledge expansion-drive”.

ii) The right-hand-side type – Suppose a kid  is shown, say, a crab or a lobster in a restaurant for the first time and it occurs to it after observing it for some time – what might be inside it? Now this is not a sheer knowledge-seeking curiosity exercise. There is a genuine thought that occurs to him about what is shown to it.

  1. Intermediate steps of Thought or Knowledge towards a Thought-goal or a Knowledge-goal  : 

Illustrations of the above 3 types – 

i) The fist-type – Here the goal is a certain pre-known knowledge piece, and there is a thought-based question as an intermediate step towards that goal.

E.g. – How much protein do I have to consume per day for the next 8 months, if I have to increase my weight from 50 kg to 70 kgs? This is a thought-based intermediate question-step towards a pre-known goal of 70 kg weight from 50 kg weight in 8 months. 

ii) The second-type – Here the goal is a certain piece of thought-inference, and there is a knowledge-based question as an intermediate step towards the goal.

E.g. – Suppose someone tells you that his friend is 6 feet 5 inches tall. And then you immediately ask him about his friend’s weight. Here, you ask this intermediate knowledge-based question to infer about his body-constitution/fitness (weight : height ratio).

iii) The third-type – Here the goal is a certain piece of inference, and there is a thought-based question as an intermediate step towards the goal.

E.g. – Is this much decoration enough? Here the immediate question is a thought-based question/judgment about an amount (amount of decoration) as to a thought-based target of it being judged as enough/good enough.

Naturally suitable data-structures for cognition

Naturally suitable data-structures for cognition –

  • connections (sometimes arrows) emanating a point.
  • chains / flowcharts
  • enlisting / enumeration
  • classification    
  • etc.

These cognitively convenient organisational formats match with the Linguistics (simply speaking, the language) of our first reaction in trying to understand / grasp / get a hang of any piece of new information given to us. That is, even without any pen or paper.

They are the “natural” formats of cognition – try to recall how you say to yourself, slowly and silently chanting, bit by bit, the parts of any new information being presented to you, in an attempt to ORGANISE things while and thereby trying to make sense of the information in your mind

E.g. of these above-mentioned first-reaction “chantings” – the language of our attempt to understand – 

“OK. So there is a tiger, a goat and a dog” – Enlisting.

“OK. So from here the water goes there, from there to there, and from there to there” – Chain.

“OK. So there is a Mr. Smith, he has 4 brothers. Each of them has 2 daughters” – Connections emanating from a point and further – tree diagram.

“OK. So there are 3 kinds of mammals. Examples of the first kind are X and Y, examples of the second kind are A and B …..” – Classification.

Cognizing a Question

Consider these 2 points – 

1) There is something which can be called “Linguistic-tapping”. Brief explanation – When questions come, they come in the form of a few words. And our mind-system has reservoirs of inter-connected knowledge-bases which are tapped by these few words (Linguistic pointers to the databases/reservoirs) while thinking, to answer the given question. 

2) Also, there is familiarity and hence a “comfort” with certain words since we have huge respective corresponding reservoirs owing to familiarity with multiple ways, forms and contexts of usages of those words. For example, consider words like ‘good’, ‘man’ etc. (occurring in questions – the context in concern here). These words are so familiar and have been heard so many times, being used in varied senses and contexts, that there is an automatic “comfort-zone” with these words.

The above 2 points collectively imply the following : Whenever there is a question posed to us, the “familiar” parts get quickly “set and organised” into the mind, with the unfamiliar, odd & less-heard-of words being are held actively by the mind and are harped upon while thinking about that question. 

This is also supported by the following observed fact : 

Consider these 2 examples of questions – 

1. Which would be the fastest car on Mars?

2. Who is the smartest man in space?

In these, the fastest car (1st question) and the smartest man (2nd question) are the familiar parts, and which get set and understood and organised immediately upon hearing the questions. Whereas the mind keeps harping upon ‘Mars’ & ‘space’, being puzzled by their occurrence. So, for example, the initial thinking process (in the first question) would comprise of questions like – “Mars?” / “on Mars?” / “why Mars?” / “what so special about Mars?” /….. and other such Mars-centric questions.