expectancy and familiarity

But the preference for familiarity has clear limits. People get tired of even their favorite songs and movies. They develop deep skepticism about overfamiliar buzzwords. In mere-exposure studies, the preference for familiar stimuli is attenuated or negated entirely when the participants realize they’re being repeatedly exposed to the same thing. For that reason, the power of familiarity seems to be strongest when a person isn’t expecting it.

The reverse is also true: A surprise seems to work best when it contains some element of familiarity. Consider the experience of Matt Ogle, who, for more than a decade, was obsessed with designing the perfect music-recommendation engine. His philosophy of music was that most people enjoy new songs, but they don’t enjoy the effort it takes to find them. When he joined Spotify, the music-streaming company, he helped build a product called Discover Weekly, a personalized list of 30 songs delivered every Monday to tens of million of users.

The original version of Discover Weekly was supposed to include only songs that users had never listened to before. But in its first internal test at Spotify, a bug in the algorithm let through songs that users had already heard. “Everyone reported it as a bug, and we fixed it so that every single song was totally new,” Ogle told me.

But after Ogle’s team fixed the bug, engagement with the playlist actually fell. “It turns out having a bit of familiarity bred trust, especially for first-time users,” he said. “If we make a new playlist for you and there’s not a single thing for you to hook onto or recognize—to go, ‘Oh yeah, that’s a good call!’—it’s completely intimidating and people don’t engage.” It turned out that the original bug was an essential feature: Discover Weekly was a more appealing product when it had even one familiar band or song.

- The Four-Letter Code to Selling Just About Anything (Derek Thompson)

Familiarity does not only breed trust but it also facilitates understanding. It is a well known fact that a completely new product category is very hard to market. In other words, too much innovation can be a hard sell. You need to anchor people down by creating analogical bridges to familiar categories.

But too much familiarity is dangerous in the sense that a familiar object has a greater tendency to blend into the background. (See previous post on the irrelevance of constancy.) You need to embed the familiar inside the unfamiliar to reawaken a response.

büyüme hırsı ve karaktersizlik

Yatırımcılar büyüme görmek isterler. Vizyon çok da umurlarında değildir. Rakibiniz büyüyor mu, şu an sizin de hemen büyüyüp ona yetişmenizi beklerler.

Twitter'ın ölümü bu baskıdan ötürüdür. Facebook gibi herkese hitap etmek adına ürün zamanla karaktersizleştirildi ve esas kullanıcılar platforma küstürüldü. Oysa Twitter sosyal medyanın belki de en değerli ve eğitimli kitlesine hitap ediyordu.

Sosyal ağlar da insanlar gibidir. Herkesi mutlu etmeye çabalayan biri hakkında ne düşünürsünüz? Ben şahsen karaktersiz olduğunu düşünürüm.

Snapchat'in çizgisini bozmamasını, yetişkinlere de hitap edeceğim diye çırpınmamasını son derece vizyoner buluyorum. Uzun dönemde sadece bu şekilde kitleleriyle ve duruşlarıyla ayrışan platformlar ayakta kalacaklar.

müdavim dolabı

Bazı lokantalarda müdavim müşterilerin arta kalan rakılarını saklayabilecekleri şeffaf buzdolapları sunuluyormuş. Herkesin görebileceği yerlere konan bu dolaplarda, şişelerin üstlerine isim etiketleri yapıştırılıyormuş.

Ne kadar zekice bir buluş!

1. İsraf engelliyor, arta kalan rakılar çöpe gitmiyor.
2. Aitlik hissi ve bağlılık yaratılıyor. Müşteriler mekanda kişisel bir iz bırakıyor ve geri dönüp kalan rakılarını bitirmek istiyorlar.
3. Daha fazla rakı tüketiliyor. Artsa da saklayabilecekleri için, müşteriler korkmadan küçük yerine büyük rakı açıyorlar.
4. Normal müşteriler bu dolabı görünce onlar da müdavim olma hevesine kapılıyor. Daha sık restoranı ziyaret ediyor ve sonunda onların da isimleri müdavim dolabına giriyor.

Dizayn konusu ilginizi çekiyorsa, şişeler ve bombeler üzerime yazımı da okuyabilirsiniz.

dikkatsiz veletler

Öğretmenlerimiz şikayetçi. Yeni jenerasyon çok problemliymiş. Öğrencilerin genelinde dikkat sorunu varmış.

Çözüm: Çocuklara ilaçları dayayalım, dikkat artırdığı vaat edilen saçma sapan oyunları zorla oynatalım.

Kimse düşünmüyor nasıl eğitim içeriğini daha ilginç kılabiliriz, nasıl eğitim formatını içeriğe uygunlaştırabiliriz, nasıl çocukları motive edebiliriz diye.

Kimse şüphelenmiyor öğretmenlerimizin kalitesi düşüyor, internet üzerinden akıcı eğitimsel içeriklere kolayca ulaşabilen öğrencilerin dandik öğretmenlere tahammülleri azalıyor diye.

safety of extremes

The great artists and the bad artists are easy - it is the good artists that can kill you. With the great artists you just keep putting fuel in their tank. With the bad artists, you realise your mistake quickly and cut your losses. It is the good artists that bankrupt you because they are good enough to make you think they are about to turn the corner and therefore keep you spending.

Blockbusters - Anita Elberse (Page 81)

Similarly, you do not have to be afraid of stupid or smart people. You will feel tempted to set up mechanisms to keep the stupid ones in control and you will confident delegating responsibilities to smart ones without any worries.

Those in the middle will constantly fool you and turn out to be the costliest.

3 pillars of risk analysis

At Urbanstat, our philosophy of risk analysis is all-embracing and rests on three complementary pillars each of which has its own upsides and downsides.

 

Statistical Modeling

Generally speaking, risk analysis has always been about deciphering statistical patterns. What has changed over time is the sophistication of the models employed. Simple linear models have been discarded in favor of ensemble models that combine different types of approaches and go beyond the traditional least square estimation techniques.

Hence, in some sense, the modeling community has embraced the values of the post-modern world where no approach is deemed to be inherently correct. Every approach has its own unique context-dependent set of advantages and disadvantages.

As Urbanstat, we use ensembles consisting of decision trees and neural networks to help insurers detect the high-risk customers. Since we only know the fate of the accepted policies, we can warn the underwriters only about risks that they are willing to accept but should not. In other words, statistical modeling cannot warn about false negatives, policies that are being rejected but should not. Despite this fact that we can only see one side of the moon, we can still create enormous value for our clients, helping them see the complex statistical patterns that go unnoticed.

Models are tailor-made for each of our clients. We clean and enrich the data sets, supervise the variable and model selection processes. We work closely with our clients to ensure that the resulting decision-making assistance suits their risk appetite.

Downsides:

• Cannot detect false negatives
• Cannot provide humanly comprehensible reasons for rejection

Upsides:

• Unlocks humanly incomprehensible complex patterns
• Improves continuously over time

Physical Modeling

Unlike most other types of risks, due to their mechanical physical nature, geographical risks can be gauged even in complete absence of past policy/claims data. In this sense, Urbanstat’s geographical focus has provided it an important fallback option when statistical analysis is not feasible.

Catastrophe modeling is hard because catastrophes are both complex and rare. We either import external models or develop our in-house ones if we believe that we can do a better job than the existing alternatives.

Our ultimate vision is to become completely model agnostic by establishing a marketplace where institutions (companies, universities etc.) can put up their catastrophe models for sale. After all, as in the ensemble approach to statistical modeling, conjunctional use of different physical models often improves the outcomes.

Downsides:

• Cannot be updated very frequently
• May have a high margin of error depending on the complexity of what is being modeled

Upsides:

• Can help the underwriter even in complete absence of past policies/claims within the region concerned
• Helps build further human intuition via visual layers

Human Intelligence & Institutional Policies

Although there are talks of complete automation of underwriting services, we believe that it will not happen anytime soon. Machine intelligence and human intelligence work in different ways and each have their own advantages. That is why the hybrid approach always performs better, even in very well-defined contexts like chess games.

Moreover, one should never forget that it is the humans that provide the data sets that machine learning algorithms get trained on. Hence there is always a continuous need for human inputs.

In Urbanstat, we allow underwriters to easily draw authorization regions and add flexible if-then rules on these regions. Through this general mechanism, they can incorporate into their risk analysis framework all the institutional policies and individual insights.

Downsides:

• Subject to human and organizational biases
• Can get complex to manage and monitor as the underwriter team scales

Upsides:

• Adds anticipative power to the whole framework
• Improves statistical models that feed on human decisions

diversification reflex

Risk selection is not about blind diversification, it is about right kind of diversification.

At Urbanstat, once we geocode our clients' policies, the resulting geographical visualisation dazzles the risk managers who have been thinking in the tabular format for years. Rightly so, they feel as if they have been blind for years. 

Their first reaction often has to do with how the company can finally handle their diversification goals more accurately. Now that they can see everything on a map, they can modify their sales goals to create the perfectly-uniform geographic distribution they have been after.

Of course, this uniformity business is exactly the opposite of Urbanstat's thesis. The whole point of our geographical approach is to bring out the unseen non-uniformities and help insurers adjust their portfolio allocations accordingly.

Blind diversification works well only after all the known unknowns are factored out. Left with the remaining unknown unknowns, there is in fact nothing to do but to distribute all the bets evenly.

Everything else being equal, the density of bets in a certain region should be lesser than the one in a less risky region. After all why assume greater risk for the same price unless all the sale opportunities in the less risky region are exhausted?

socialness, consciousness and smartness

Degree of Socialness

While humans can easily handle fourth-order intentionality chains like "I think that X thinks that Y thinks that Z thinks something", nonhuman primates seem to be capable of handling only first or second-order intentionality chains. This difference is thought to correspond to our greater social skills. (No wonder why “four” is often taken as an optimum number of active characters in any given movie scene. Spectators desire to be stimulated to the upper bound of their capabilities.)

The degree of mind-simulation capability determines the size of the Dunbar's number, which for humans is 150 and defined as the "cognitive limit to the number of people with whom one can maintain stable social relationships—relationships in which an individual knows who each person is and how each person relates to every other person."

Some people are apparently even capable of handling sixth-order intentionality chains. In other words, they are both physically and mentally six degrees away from any other person in the world. (i.e. They can mentally simulate their entire social network one-intentionality-chain at a time.)

Children start to demonstrate theory of mind at roughly around the same time that they start to recognise themselves in the mirror. “You have to be aware of yourself in the first place in order to begin to take into account what other people may know, want, or intend to do,” Gallup says. He notes that people with schizophrenia often cannot recognise themselves in the mirror, and they struggle with theory of mind as well.

What Do Animal See in a Mirror? - Chelsea Wald

We understand others via empathy which is the ability of put oneself in other's place. Hence, it is not surprising that you are able to recognise yourself in the mirror only once you are able to formulate first-order intentionality chains. (You are literally projecting yourself onto your reflection.)

In the time between the original Homo species and ourselves, the brain doubled in size. A disproportionate share of that growth occurred in the frontal lobe, and so it stands to reason that the frontal lobe is the location of some of the specific qualities that make humans human. What does this expanded structure do to enhance our survival ability to a degree that might have justified nature's favouring it? ...In addition to regions associated with motor movements, the frontal lobe contains a structure called prefrontal cortex. "Prefrontal" means, literally, "in front of the front" and that's where the prefrontal cortex sits, just behind the forehead. The prefrontal cortex is responsible for planning and orchestrating our thoughts and actions in accordance with our goals, integrating conscious thought, perception and emotion; it is thought to be the seat of our consciousness.

Subliminal - Leonard Mlodinow (Page 102-103)

The fact that the size of a species' neocortex as a percentage of its whole brain is correlated with the size of its social group implies that the relationship between self-consciousness and intentionality chains extends to higher orders too.

By the way, isn't it poetic that we feel a tender desire to touch our foreheads with the loved ones? We are literally trying to merge our consciousnesses!

Degree of Consciousness

Since consciousness is just a model of the brain itself (as pointed out in the previous blog post), we expect the volume reserved to consciousness (presumably the neocortex) to grow at the same rate as the volume of the whole brain (what is being modelled). Relative increases in the size of the neocortex as a percentage of the whole brain could be due to improvements in the fidelity of the models themselves. 

Hence, one could define "higher degree" of consciousness as any of the following equivalent statements:

• Higher fidelity cognitive models

• Less information being lost during the cognitive modelling processes

• Higher number of nested past selves one can be cognisant of at any given moment in time

Armed with this definition, the conclusion of the previous section can be rephrased as follows: The more "copies" of ourselves available to us at any given moment in time, the deeper we can simulate other minds. In other words, conscientiousness and consciousness are in some sense the same thing.

Degree of Smartness

Remember the old blog post on empathy and truth?

There we claimed that empathy allows one to get closer to truth since understanding takes place through causal statements like "A -> B" and these statements can be internalised only by literally putting ourselves in places of whatever A and B are. In other words, mental simulation of inanimate phenomena uses the same principle as that of social phenomena.

The aim of science is not things themselves, as the dogmatists in their simplicity imagine, but the relations among things; outside these relations there is no reality knowable.

- Henri Poincare

What the great mathematician Poincare is saying here is that all understanding is relational. A and B literally have to be taken as black boxes. The only thing that we can probe is the relationship between them as depicted by the arrow sign "->". (Note that this is the essence of Category Theory.)

In some sense, the conscious self is the most canonical black box at our disposal. (We can not peek into the pronoun "I".) By projecting ourselves onto A, we temporarily replace A with "I" to gain an understanding of A's relationship with other objects.

Summing up, the degrees of all of the following are correlated via the notion of empathy:

• Socialness - as defined by the experimentally-measurable maximum-length of intentionality chains one is capable of simulating

• Consciousness - as defined by the experimentally-inaccessible maximum-number of nested past selves one can be cognisant of at any given moment in time

• Smartness - as defined by the experimentally-measurable maximum-length of causality chains one is capable of simulating

domains of cognition

Did you know that emotions correspond to certain bodily states which precede the actual experience of emotions? (Read this interview with Lisa Feldman Barrett) 

Similarly, the instructions we send back to the body upon feeling a certain emotion embark on their journey before we become conscious of them.

The complete correspondence between physical phenomena and cognitive models is as follows:

• Environment <-> Perceptions
• Body <-> Emotions
• Brain <-> Consciousness

By definition, modelling involves reduction in information content. Just like we can not perceive our environment at its entirety, we can not be conscious of every single activity going on inside our brains. (Remember that evolution optimises for survival, not understanding.)

The discovery of the unconscious was traumatic. Similarly, we resisted the idea that there could be stuff out there that lie beyond our perceptions. (e.g. micro organisms, atomic particles, electromagnetic waves) Each such traumatic cultural acceptance process was followed by an outburst of mesmerisation and imagination. A grand belief in mystery reemerged and many speculative phenomena got ascribed to the newly discovered inaccessible realms.

---

The cognitive models exhibit nestedness, just like the physical phenomena they model. But the order of nestedness is inverted and the relationships are mediated via causality rather than spatiality.

• Environment > Body > Brain
• Perceptions <- Emotions <- Consciousness

Perceptions are affected by emotions. The domain of attention changes as the emotional state does.

Both emotions and perceptions are affected by the states of consciousness. For instance, you experience a lot more stuff when you are awake than when you are deep asleep.

---

You may be wondering how a brain can model itself. Would that not amount to creating a recursive loop? The model of the brain is part of the brain and therefore it too needs to be inside the model. But how can a model be inside itself?

In the timeless world of mathematics, recursions instantly turn into monstrous creatures. But in the world of physics, recursions take place in time and their behaviour get tamed.

A model of the brain at time t contains a model of the brain from time t-1. In other words, consciousness is like a Russian matryoshka doll which has (due to the enormous information loss happening at each step of modelling) a very small number of nested units.

extremity of randomness

Hell is not the most tormenting space. Limbo is.

Uncertainty is unbearable for the human psyche. Torture methods that involve randomisations are the worst. For instance, releasing water droplets onto someone's forehead at random intervals apparently drives people insane. (Disordered raindrops have a calming effect on rough oceans. It has the opposite effect on brain waves since our natural resting state itself is actually pretty wavy.)

Reward mechanisms also perform best when they involve randomisations: 

Whether the subject is a pigeon, rat, or person, Skinner found, the strongest way to reinforce a learned behaviour was to reward it on a random schedule.

- How Designers Engineer Luck Into Video Games (Simon Parkin)

In other words, randomisation has an overall amplification effect, making the negative more negative and positive more positive.

---

Although we are not good at psychologically guarding ourselves against randomised suffering, we are very good at offloading our psychological suffering onto random factors. (For instance, we consistently underestimate the role of chance in our successes and overestimate it in our failures.)

We can not offload the pain associated with randomised suffering back to random factors because stories can be deformed only after the fact, not when they are unfolding in realtime.