Page 163 - Data Science class 10
P. 163

This partiality, preference and prejudice towards a set of data is called as a Bias. Data bias leads to biased models,
            which may be unfair and harmful to people. A thorough evaluation of the available data and its processing to
            mitigate biases should be a key step in modelling.

            3.2.1. Partiality

            Partiality is tendency to favour unfairly. For example, we can say that society is partial, in which individuals are
            segregated and subjected to discrimination based on factors such as caste, creed, gender, and other characteristics.

            3.2.2. Preference
            Preference is the act of selecting or having a particular preference for one person or thing rather than another or
            others. Some common synonyms of preference are alternative, choice, election, option, and selection.
            While all these words mean "the act or opportunity of choosing or the thing chosen," preference implies making
            a decision based on one's judgement or preferences.
            In psychology, economics and philosophy, preference is a technical term usually used in relation to choosing
            between alternatives. Buyers are showing a preference for electric cars.

            3.2.3. Prejudice and/or Bias

            Prejudice means preconceived opinion that is not based on reason or actual experience. The word comes from the
            Latin "pre" (before) and "judge". They can conclude that they don't like them because of their skin colour (this is
            "racial prejudice"), religion (religious prejudice), nationality, and gender.
            Prejudice is an opinion held against a person or group that is often unfavourable and/or intolerable and is based
            on a lack of information. The word bias has very similar meaning but not as extreme as prejudice. Someone who
            is biased usually refuses to accept that there are other views than their own.

            Prejudice can be classified into three different categories: cognitive prejudice, affective prejudice, and conative
            prejudice. Some of the most well-known types of prejudice include:
               • Racism                                             • Homophobia
               • Sexism                                             • Nationalism
               • Ageism                                             • Religious prejudice
               • Classism
               • Xenophobia (fear of strangers) is a broad term that may be applied to any fear of someone who is different
              from us
               • An unfavourable opinion or feeling formed beforehand or without knowledge, thought, or reason
               • Any preconceived opinion or feeling, either favourable or unfavourable
               • Unreasonable feelings, opinions, or attitudes, especially of a hostile nature, regarding an ethnic, racial, social, or
              religious group
               • A law that operated to the prejudice of the majority and may cause damage or injury; detriment
            The difference between prejudice and partiality is that prejudice is a harm, a damage while partiality is tendency
            of preference or bias in favour of someone over the other.
            An example of prejudice is having a negative attitude toward people who are not born in the United States. Even if
            those who hold this bias don't know all foreigners personally, they dislike them because they are foreigners.

            Prejudice frequently consists of unfavourable emotions, stereotyped ideas, and a propensity to treat members of a
            group differently. In society, we often see prejudices towards a group based on race, sex, religion, culture, and more.

            Statistical bias is anything that leads to a systematic difference between the true parameters of a population and
            the statistics used to estimate those parameters.
            For example, a high school student sample will be skewed in a poll to determine teen drug use because it does not
            include home-schooled students or dropouts. A sample is also biased if certain members are under-represented
            or over-represented relative to others in the population.


                                                                                          Identifying Patterns   161
   158   159   160   161   162   163   164   165   166   167   168