msleonor-blog

[msleonor-blog]

O estranho

Tal como o indivíduo tem o direito e obrigação moral de prosseguir a sua existência, o mesmo acontece para grupos, pois que a prossecução da sua natureza só é possível existindo. Estes são definidos por uma identidade comum partilhada pelos seus membros que, embora por vezes pouco clara, é real, saliente e relevante. Desta maneira, a panóplia institucional que criam para a sua auto-preservação está prima facie justificada. Este raciocínio aplica-se sobretudo no que diz respeito à introdução de membros alheios no seio do grupo. Estes não partilham da identidade que o define, logo serão sempre estranhos. São tolerados na medida em que a sua identidade não ameace a do grupo. Outra alternativa, não mutuamente exclusiva com a anterior e apenas caso possível, é o compromisso de a abandonar total ou parcialmente em prol daquela do grupo. Pelo contrário, nos casos em que a identidade do estranho perfilha uma noção de oposição vis a vis a dominante, a integração não é possível. Aqui, o estranho define-se contra o grupo que o acolhe. E, como a identidade do grupo se manifesta de forma ubíqua para quem está no seu seio, a do estranho torna-se especialmente saliente, que a irá sobrepor à do grupo. Assim, o estranho não consegue evitar atentar contra a identidade maioritária, seja através da cadeia infinita dos porquês ou da normalização forçada das suas crenças. Fá-lo tanto de forma consciente como inconsciente, minando o grupo e a sua moral. A partir deste momento, deixa de ser justificável tolerá-lo, já que o todo está em causa.

#estranho #mine

[msleonor-blog]

O alienado

O alienado ama com a razão. O grupo e tudo o que este representa carece de justificação, um antro de hipocrisias e arbitrariedades. Ao não distinguir e reduzir o outro ao princípio, elimina sucessivamente cada e toda a parte da sua essência, até que nada mais resta para amar. No entanto, ser humano é querer amar a essência do outro; é um fim em si mesmo, pelo que o alienado cai sempre na contradição. É dele que nasce o caos, em que tudo é e não é tudo e nada ao mesmo tempo, pela total redução do fim social a uma mera relação contratual acidental de benefício mútuo. Por isso, o alienado corrói e destrói, tornando-se aquele que acusa e divide.

#mine #evil #alienado

[msleonor-blog]

O segredo

A insignificância dos nossos dias pesa-nos na alma, mas é neles que se constrói a nossa história. A cada palavra, a cada gesto damos mais um passo para a realização da nossa essência. Ser humano é estar ciente disto. Ser feliz é abraçá-lo. No dia em que amamos por inteiro o pequeno, adormecemos com um sorriso. Não existe a cadeia dos porquês, não existe a suplantação do divino; apenas o apego ao que nos é próximo e pequeno - como nós.

#mine #comfie #love

[msleonor-blog]

Thank you

I want to thank you today. I want to thank you because during these years you did not let the child inside me die. Regardless of how my day has been, in the end, I find you running towards me with a welcoming smile. That warmth fills me and suddenly all I want to do is smile - all I want to do is make everyone around me happy.

I love the little duets we make when we whistle together. I love the way you eat bread and cookies like a little mouse and get mad when you realize I am noticing that. I love to use you as my little guinea pig for the hairstyles I make up, and how cute you always look in the end. I love, even though I never understood why, how you feel you must have to jump up and down the stairs, like a penguin would. And you make me do it too! I love how you tilt your head a little and smile too much when you know you have screwed up, and you never understand how I can always guess you did something wrong. I love how we make a mess of ourselves and everything around us whenever we try to create something together, like cooking or painting. I love the little plays we make whenever I am trying to read you a book. I love how calm and peaceful you look when you fall asleep, and how you snuggle close to me if I’m close. I love how happy you are when you wake me up by tickling my feet. I don’t like it very much but your little burst of happiness is well worth the price. I love your endless whys, the way you jump a little when you understand something, and being surprised with how smart you are. I love how you treat your little one (shh!) the same way I treat you.

I love how you bring me hope for tomorrow, how you make me wish to be a mother in the future, and how watching and contributing to raise a happy child makes me feel.

But, most of all, I love you. Thank you.

#mine #love #family #children #comfie #warm #thanks

[msleonor-blog]

Taste discrimination - II

An important question remains to be answered about taste discrimination: why does it exist? Why do we prefer to interact and are more sympathetic towards people who are a similar to us on a vector of characteristics?

One popular answer is social conditioning; that is, socially constructed characteristics are made salient by our social network and the media and, through exposure or simple social pressure, we grow to prefer people who possess the valued characteristics made salient. However, there are two major problems with this explanation. The first is that it does not make clear why some characteristics are made salient in the first place. The second, more serious, is that it provides no clue as to why in-group favoritism is found in cross-cultural studies, which suggests it is a universal human trait. Universalities of a trait often suggest that the explanation must rely on an evolutionary foundation.

One possible explanation focuses on scarcity. In the context of scarcity, groups whose individuals favor themselves over others will tend to do better compared to groups whose individuals don't. As an example, consider an iterative process with agents who belong to a given group – with observable characteristics - and follow one given strategy. The game being played is one-shot prisoner's dilemma to rule out reciprocity effects. The final outcomes enjoyed by each agent affect its probability of generating other agents similar to itself. Each agent can follow four strategies: 1. Altruists: they are cooperative towards everyone, regardless of the group; 2. Egoists: they do not cooperate with anyone, regardless of the group; 3. Ethnocentrists: they cooperate towards members of the same group, but do not towards members of another group; 4. Xenocentrists: they do not cooperate towards members of the same group, but do towards members of other groups.

In this framework, some patterns are found. First, clusters of altruists do better and expand more than clusters of egoists, for all possible groups. This means that there is a positive effect to growth for being ethnocentric or altruist. Second, however, egoists do better when interacting directly with altruists, putting a drag on the growth of ethnocentric and altruist clusters. Third, and most importantly, when clusters can no longer expand, clusters of either altruists or ethnocentrists – all belonging to the same group– will start to face clusters of pure egoists. Altruists benefit a lot from having other altruists around but they are harmed for following a cooperative strategy against egoists, meaning the egoist cluster will grow at the expense of the altruist one. On the other hand, ethnocentrists enjoy the same benefits of the altruists for having other ethnocentrists around but they are not harmed by egoists in other clusters if they are of a different group. This means that the cluster of ethnocentrists – again, all belonging to the same group – will grow at the expense of the cluster of egoists. All in all, there will be a tendency for the cluster of ethnocentrist to grow at the expense of other clusters, unless they are clusters of ethnocentrists belonging to a different group. Fourth, and last, xenocentrists quickly disappear - they perform as bad as egoists when not space-constrained and when they face both egoists and ethnocentrists of other groups, even though they behave as an egoist inside a cluster of ethnocentrists of the same group. Again, note the game above does not require antipathy towards members of other groups, just that there is a difference in treatment between in-group and out-group members. Also, group membership needs be observable in some way for these results to hold, even if there is uncertainty.

#taste discrimination #discrimination #racism #sexism #ingroup favoritsm #mine #ethnocentrism #xenocentrism #altruism #egoism #evolutionary #scarcity

[msleonor-blog]

The brain you don’t control - VI

Filling in the gaps - blind spot in the eyes.

“Another popular one – and quite unnerving if you’ve never seen it before! Here’s what I want you to do: close your left eye and just focus your right eye on the tiny static circle on the left of the screen. At some point the big circle will disappear as it crosses your ‘blind spot‘. If you can’t see this effect, it means you’re sitting too close/far from your monitor. Try sitting closer/further to the screen and repeat the test. So, have you managed to find your blind spot?”

http://www.moillusions.com/find-your-blind-spot-trick/

#brain you don't control #mine #blind spot #automatic #autonomous #brain #perception #constructed reality

[msleonor-blog]

This is just heartbreaking..

#degeneracy #heartbreaking #sad #video #mine #let's go back

[msleonor-blog]

Taste discrimination - I

I have showed that discrimination can occur even in the absence of discriminatory preferences. Instead, it would occur because uncertainty forces one to rely on group information to assess individuals. A more pernicious form of discrimination is precisely the one based on discriminatory preferences. Here, unlike the previous form, more certainty actually makes it more likely that the individual is discriminated against. Most importantly, like in the case for statistical discrimination, unless only a minority is engaging in taste discrimination, competition does not necessarily eliminate it.

The framework for taste discrimination is actually very simple if compared to that of statistical discrimination. Individuals are divided into groups and evaluate others differently as a function of their group membership. The most common form is in-group favoritism, where individuals prefer interaction or inputs from those of their own group relative to others, to whom indifference is displayed. I must emphasize this point about indifference. It is often assumed that in-group favoritism entails outward antipathy. That is not necessarily the case, as empirical evidence seems to show independence between the two. Hence, most taste discrimination takes shape as one being more likely to help members of his group in ambiguous situations, more likely to give them the benefit of the doubt, less likely to interpret their actions as aggressive, etc. Again, in other words, there is a positive bias towards members from one’s group but no bias towards members from other groups.

#taste discrimination #discrimination #racism #sexism #ingroup favoritism #mine #preferences

[msleonor-blog]

Statistical discrimination - IV

The point presented in the last post is not realistic. It assumes that the only thing we know about the individual we wish to know more about is a given group membership. Most of the time, however, we also have information on a signal for the variable of interest: think about a performance test to measure the productivity of a worker. The signal is not perfect: on average, it gets things right but there are possible errors. This means that it's possible two individuals perform equally well on the signal even though there is an underlying difference in the variable of interest.

This point is better explained mathematically. We have two groups, A and B, whose distribution for the actual variable of interest, \( y \), is: $$ y_{ix} = \bar{y_x} + \varepsilon_{ix},\space \varepsilon_{ix} \sim N(0,\sigma_{\varepsilon,x}^2), \space x = \left\{A,B \right\} $$

This variable is not observed (finding its value is the objective) but its mean and variance for each group, \( \bar{y}_x \) and \( \sigma_{x}^2 \), are assumed to be known. At this point, we are in the same position as in the third post. However, there is also a signal that gives us information about the variable of interest, which, by definition, has errors: \[ y ̃_{i,x}=y_{i,x}+\eta_{i,x}\]\[\space \eta \sim N(0,σ_{η,x}^2 ),\space x= \left\{A,B \right\} \]

Hence, if we linearly regress \( y \) on \( \tilde{y}_{i,x} \) ( \( E[y_{i,x} |y ̃_{i,x},x] = \alpha_x+\beta_x \tilde{y}_{i,x} \)), we find that: \[ \alpha_x = \bar{y}_{x}(1-\beta_x) \]

\[ \beta_x = \frac{\sigma_{\epsilon,x}^2}{\sigma_{\epsilon,x}^2 + \sigma_{\eta,x}^2} \]

This implies that the expected value for \( y \) given the signal and the group, is: \[ E[y_{i,x} |\tilde{y}_{i,x},x] = \bar{y}_x + \beta_x [\tilde{y}_{i,x}-\bar{y}_x] \]

The difference between two individuals with the same score for the signal, \[ E[y_{i,A}|k,A] - E[y_{i,B}|k,B]\]\[ = \bar{y}_A(1-\beta_A) + \bar{y}_B(1-\beta_B) + k(\beta_A - \beta_B) \]\( \bar{y}_x \) and \( k \) are assumed to be positive and \( \beta_x \in ]0,1[ \) by construction.

All of this, has interesting implications:

If the two groups are equal on every regard, there will be no statistical discrimination;

If only the means of the different groups differ, the difference will be \( (\bar{y}_A - \bar{y}_B)(1-\beta) \). In the example given above, this would mean that the worker from the disadvantaged group actually needs a better score comparatively to the worker from the advantaged group to be hired;

If only the dispersion of the groups is different (consider \(\sigma_{\epsilon,B}^2 > \sigma_{\epsilon,A}^2 \)), \( \beta_B > \beta_A \), implying the lines for the expected value of the variable of interest as a function of the proxy cross. Following the example, this means that an individual from B is much more likely to be hired if he has a high score, but much less likely if he has a low score, relatively to an individual from A;

If the reliability of the tests differs (consider \(\sigma_{\eta,B}^2 > \sigma_{\eta,A}^2 \)) , \( \beta_A > \beta_B \) , implying the same as in 3. but reversing the roles;

Note that, if \( \sigma_{\eta,x}^2 \) is 0 for both groups, there ceases to be any discrimination and individuals evaluations will be based on their own merit and not on group information. This happens because now you completely observe the previously unobserved variable.

I have done a ceteris paribus analysis here but, most likely, multiple effects are going on at the same time. More concretely, consider white and black workers. White workers are more educated on average and have higher IQs, meaning that their performance on the signal of productivity is, on average, better, so blacks will be discriminated against according to point 2. We might also think that the dispersion in productivity for black workers is greater, meaning that an extra point in the signal score is better rewarded (point 3). However, there is reason to also believe that the proxy itself is more reliable for white workers, point 4, so the previous effect is counteracted. Empirical evidence seems to show that better signal scores (through education, experience or SAT scores) reward white workers (or students) better, so point 4 seems to be stronger than point 3.

This series of posts is mainly to show that there can be discrimination even in the cases where there is no distaste for a particular group. In other words, actions might show sexism or racism but thoughts do not. Instead, discrimination only occurs because individuals do not know everything about each other and rely on group information to make decisions. Do note that ignoring group information lowers the quality of the decisions. That is, if the goal is to be as efficient as possible decision-wise, considering also the point about the differences between FP and FN, then one should rely on group information. Examples of phenomena that can be explained using this framework are racial profiling, discrimination in hiring, discrimination in university admissions, discrimination in dating or prejudice against immigrants.

#statistical discrimination #information #statistics #discrimination #signal detection #stereotypes #racism #sexism #racial profiling #mine

[msleonor-blog]

Statistical discrimination - III

To decide how to act towards other individuals, we need information about them. As I have argued in the previous posts, this information is limited. Still, we do possess means of extract it through heuristics. Do note that if it were not limited, we would still use heuristics given current conditions but they would probably have never evolved in the first place. Furthermore, these heuristics might be biased or not depending on the relative costs of FNs or FPs.

We are typically interested in using the information we possess about an individual to track a given variable of interest about him. Think, for instance, that we would like to know how likely he is to cooperate with us instead of free riding on the group, or rather how useful he would be if he were to be added to a team.

Through the most salient features of the individual, we will fit him inside several categories, which carry social expectations. These are also known as stereotypes, which are nothing more than expectations on the perceived base rate probabilities of the individual's group for a set of traits. These are in many cases accurate and in others inaccurate. In fact, they need not be accurate precisely because of the heuristic bias raised in the second post.

Still, this raises a possible issue. If the individual belongs to a group with high base rates for a set of negative traits, it would be assumed that he also possesses them. In other words, he will be considered a positive. Now, he can actually be a TP or a FP. If he is a TP, everything is fine as the prediction based on limited information was correct. However, if he is a FP, he will bear costs for other people assuming he has negative traits he doesn't actually have. Yet, this does not mean that information on group membership should be discarded, as the point is to avoid not only FPs, bore by the individual, but also FNs, bore by the decision maker. In fact, this means that abstaining from using base rates will diminish the quality of our decisions, as we would have no choice but to base it on something akin to a uniform prior. Also, this raises a conflict of interest, as the judged individual would like to avoid FPs at all costs, whilst the decision maker wants to avoid FNs.

#statistical discrimination #discrimination #information #statistics #stereotype #racism #heuristic #base rate #bias #mine

[msleonor-blog]

Harm principle - IV

In the second post, my main point was to argue about indirect effects that would violate the requirements of the harm principle. That is, there is something X I value that is not directly related to individual A. However, the actions of A bring about changes in X that harm me. Think about laws banning smoking in closed spaces as an example. In such cases, the requirements to apply the harm principle do not hold and, hence, there would be justification to limit the choice of A. Now, my point here is to raise a related issue. It's perfectly plausible that I hold preferences on the actions of A, meaning that his action would bring about harm regardless of other conditions. Furthermore, I can also hold preferences on the preferences of A, so that I am harmed by realizing that A prefers what he prefers and not otherwise. Knowledge on the actions or preferences of A because we can introduce uncertainty in the mix or information can be revealed through omission. This problem exacerbates the issue of incommensurability of utilities raised in the third post and makes it much more likely that the requirements of the harm principle would not be met following the second post.

Do note that suggesting these preferences related to others are not legitimate would be going against the harm principle, as would be claiming they ought to be changed, for most cases. Also, such claims reek of arbitrariness and ad hockery.

Therefore, the fact that people have preferences over the actions and preferences over others makes the application of the harm principle much harder.

#harm principle #mine #stuart mill #liberalism #preferences on preferences #morality #ethics

[msleonor-blog]

Statistical discrimination - II

Since we cannot perfectly anticipate what's going to happen, we need to rely on statistics. The object of analysis can be known unknowns - in which we face risk - or unknown unknowns - in which we face uncertainty. To solve the first set of problems, standard optimization with expected values is enough; however, for the second, we will always have to rely on heuristics. And the latter case is exactly what we, humans, do. Risk is very uncommon in nature and that's why we are notoriously bad at dealing with it - we apply heuristics to problems of risk and we naturally get the wrong answers.

Now, any statistical analysis is bound to err. We might either say something is there where it isn't, a false positive (FP); or that nothing's there whereas it is, a false negative (FN). Obviously, the two remaining cases are true positives (TP) and true negatives (TN), where we either saying something or nothing's there and we get it right.

If our brain was equipped to truthfully assess what's going on in our environment, we would expect that the heuristics we follow minimize errors - that is, minimize the number of FP and FN. However, that's not how evolution operates. Truth, evolution-wise, is useful insofar as it aids in our survival, meaning that what really matters is the cost of committing mistakes. This implies that if, for instance, a FN is more costly than a FP, the heuristic we use will be biased towards minimizing FN. In other words, we will be trigger happy and say "something's there" when it's not much more often than reasonable. This is why we misinterpret many movements in the dark - getting a FN here used to be fatal, whereas a FP is nothing more than a little scare.

#statistical discrimination #statistics #information #discrimination #snow white #false positive #uncertainty #heuristics #mine #video

[msleonor-blog]

Statistical discrimination - I

Information is costly both to acquire and process. Many times, even, this cost is impossible for one of these tasks or either. Therefore, it's not surprising that virtually all our decisions are taken with less than the full amount of information we could have. Often, even, we have none and don't have a clue what we are doing. Still, we cannot refrain from taking decisions, so we have learned (or rather evolved) to live with uncertainty: we have some clue about what's happening but the rule is unknown unknowns. Fortunately, our brain is equipped with three amazing tools that allow us to: 1. Categories: notice similarities between objects/events and make categories from them; 2. Correlation: notice how two categories (think of events) seem to follow each other; 3. Interpreter: make up plausible stories about the previous point.

This set of tools is extremely powerful but there are reasons to believe it could go wrong, namely in the interaction between the second and the third. Correlation does not guarantee causation and whatever story we make will be one of causality. Yet, it will be built upon correlations.

Despite this, these three points give us amazing advantages. The first two greatly simplify the world and identify relationships between objects/events. Even if we identify a correlation and can say nothing about causation, a statistical regularity is useful information when making decisions, if we are small not to affect the system. As for the third, it will actually make us slowly converge to the true data generating process, as we are very sensitive to contradictions. Humans have a tendency to see patterns, and namely agency, whether they exist or not and we are motivated to explain them even when we only lose when doing it - pure (known) random events.

#information #discrimination #statistics #categories #correlation #causation #patterns #inequality #differences #interpreter #statistical discrimination #mine

[msleonor-blog]

Tactical nihilism #3: claim the other person is being duped by someone, so you trigger a cheater aversion in her and make her less confident in her views.

#tactical nihilism #cheater #cheater detection #duped #conspiracy #nihilism #trust #betrayal #mine #rethorical tricks

[msleonor-blog]

Constants in the universe

There are three constants in this world: scarcity, change and inequality.

Scarcity: because of our nature as biological beings with an upper bound on the total energy of the universe and a lower bound on the total energy available on Earth. If we were to grow a 1% every year, starting with the current population on Earth, it wouldn't take more than 17000 years for us to occupy the whole observable universe, a bleep in the time scale of the universe.

Change: even though energy is finite, energy conversion is contantly happening, which creates most of what we see in the world. Furthermore, heat itself is movement.

Inequality: dictated by scarcity and change. Scarcity makes it that two objects will never share the same properties, for at least position is rival. On the other hand, change will mean that, as time progresses, the history of two objects will become more and more different.

#universe #inequality #scarcity #change #constants #mine

[msleonor-blog]

Harm principle - III

Ruling out the unlikely possibility of an individual's action having no effect on other parties (which already partly deviates us from the harm principle) and assuming we live in a world where we could measure this effect with a great deal of certainty, it could be argued that an utilitarian position should be adopted for practicality. Under this scenario, if my action greatly benefits me but brings a small harm to the other party, it should be allowed for there is a net benefits when both parties are taken as one. Still, this runs into the wall of requiring comparable utilities to work. Any interpretation we might take of utility, be it ordinal or cardinal, we will find no clear justifiable way of comparing the benefits and harms of all parties. Above, I assumed each would count equally but it is easy to find examples based on several measures (e.g. wealth, degree of need, physical differences) which would go against our moral intuitions. The same can be said for the cases where individuals are pondered differently. Hence, it can never be properly established that my action, if it benefits and harms another party, has a net benefit for both. The only way to escape this is to allow the parties to negotiate amongst themselves to achieve a Pareto efficient outcome. Yet, negotiation is an action costly in itself, leaving many effects unattended depending on their magnitude and distribution amongst the parties. Also, rights have to be clearly defined and they are not, as that is exactly what we are trying to do in the first place. Therefore, even under certainty, we cannot properly compare the harms and benefits of our actions on other parties.

#harm principle #stuart mill #utility #liberalism #utilitarianism #consequentialism #morality #ethics #mine

[msleonor-blog]

The brain you don’t control - V

Perception vs. encoding

#brain you don't control #automatic #autonomous #brain #stroop effect #perception #encoding #mine