Note. Top three responses were indicated.
Utilizing those items (Tables 2-4), several large-scale surveys in Japan were designed or analyzed by psychologists or other academic researchers after 1990, which clearly confirmed self-reported personality traits using ANOVA (Matsui, 1991; Muto et al., 2012; Ryu & Sohn, 2007; Sakamoto and Yamazaki, 2004; Yamaoka, 2009; Yamazaki & Sakamoto, 1991). Archetypical sample sizes of these were 6,660 (Yamaoka, 2009), 11,766 (Matsui, 1991), 32,347 (Sakamoto and Yamazaki, 2004; Yamazaki & Sakamoto, 1991 & 1992 – these three used the same sample) and over 100,000 (Muto et al., 2012; our estimation; the exact number was not specified in this report, although it alluded the size was much larger than preceding ones). These results cannot be explained by the Burnham effect (vague and general descriptions), because the differences in self-reports were not either virtual, vague, nor ambiguous, but supported by definite and real values.
Regardless, statistical differences had not been confirmed in respondents without blood type personality knowledge. Therefore, the current scientific consensus is that these differences are self-fulfilling phenomena induced by the “contamination by knowledge” (Cramer & Imaike, 2002; Eysenck & Nias, 1982; Kim, Lee & Lee, 2007; Matsui, 1991; Ryu & Sohn, 2007; Sakamoto & Yamazaki, 2004; Sato & Watanabe, 1992; Yamaoka, 2009; Yamazaki & Sakamoto, 1991 & 1992).

Reliability and Validity of Single- or Two-item Scale

In recent years, scales that attempt to measure psychological constructs with a very small number of items had been developed in various fields, such as scales to measure subjective well-being (Diener, 1984) and self-esteem (Robins, Hendin, & Trzesniewski, 2001) with a single item. Both scales are used in many studies. Similarly, in the Big Five personality test, a very brief measure, 5 and 10-item inventories were developed. For example, Gosling, Rentfrow and Swann (2003) created the Ten Item Personality Inventory (TIPI), which measured the five factors of the Big Five in 10 items, 7-point scale. Later, Oshio, Abe and Cutrone (2012) developed a Japanese version of Ten Item Personality Inventory (TIPI-J). The results of multiple validation tests generally supported the reliability and validity of the two tests: the Cronbach’s alphas were 0.40-0.73 (Gosling, Rentfrow & Swann, 2003) and 0.72-0.91 (Oshio, Abe & Cutrone, 2012); test-retest reliabilities were 0.62-0.77 and 0.64-0.84, respectively. Furthermore, according to several studies of psychologists (Cho et al., 2005; Ryu & Sohn, 2007; Wada, 1996), more than ten items of the Big Five scales of personality trait correspond to blood type personality (Appendix B).

Results of Personality Tests

There are many types of personality tests used in psychology, depending on the purpose. The “Big Five” test is generally used for blood type and personality studies. The Big Five personality test, as the name implies, comprehensively describes personality by five factors called the Big Five (Goldberg 1990 & 1992). These five factors are usually called Neuroticism, Extraversion, Openness, Agreeableness, and Conscientiousness. The model collects vocabularies from dictionaries and traditional personality tests, as well as re-analyses of personality scales, and five factors were extracted through factor analysis. Thus, the Big Five does not assume any background theory. In other words, it can be said that the Big Five model was constructed as the result of an attempt to broadly describe personalities with as few factors as possible, without assuming any background theory. Thus, the Big Five models are characterized by the bottom-up process and personality is comprehensively captured by five factors. The NEO-PI-R, commonly used as a Big Five personality test, consists of 60-240 question items, each of which is rated using a five-point scale (Costa & McCrae, 1992; Kunisato, Yamaguchi & Suzuki, 2008; Shiinonaka, Nakazato, Gondo, Takayama, 1998; Wada, 1996).
As mentioned above, the Big Five is a questionnaire-based personality assessment, which consists of answering to multiple questions regarding multiple self-reported personality traits. These traits integrate into five personality factors by statistical processing. In theory, this means that the self-reported answer will either directly or indirectly appear in the result. Although there are many academic studies using personality tests (including the Big Five) on the relationship between blood type and personality, the inconsistency among results (Cattell, Boutourline & Hundleby, 1964; Cho, Suh & Ro, 2005; Cramer & Imaike, 2002; Flegr, Preiss & Klose, 2013; Furukawa, 1927 & 1930; Gupta, 1990; Jogawar, 1983; Kim et al., 2007; Lester & Gatto, 1987; Mao, Xu, Mu, Ma & He, 1991; Nawata, 2014; Rogers & Glendon, 2003; Sharifi, Amadian & Jalali, 2015; Sato & Watanabe, 1992; Shimizu & Ishikawa, 2011; Wu, Lindsted & Lee, 2005) has led to the endless academic controversy about whether the relationship is scientifically confirmed. As another example, a 2014 10,000-particpant study in Japan and the US re-analyzed data of large-scale social surveys on money and consumer life revealed no meaningful difference (Nawata, 2014).
After 2000, a growing number of studies proved the previously questioned link between blood type and physical constitution, with the exception being in the weak gastrointestinal tract: this demonstration proposed a new approach to medicine (Ewald & Sumner, 2016; Risch, 2000). For example, more than 10 studies had reported a relationship between susceptibility to COVID-19 and blood type (Barnkob et al., 2020; Ellinghaus et al., 2020; Hoiland et al., 2020). According to these studies, type O was the least susceptible, and type A the most.
There had also been several studies on biological factors, which investigated whether physical constitution affected personality (Hobgood, 2011a & 2011b). In 2015, a genotype of blood type and the personality using the Temperament and Character Inventory (TCI) had been determined to be related, as predicted by blood type personality theory (Tsuchimine, Saruwatari, Kaneda & Yasui-Furukori, 2015). In this study, type A was found to be the most “persistent.” The TCI, a top-down personality model, is often used to examine genetic dispositions (Cloninger 1987; Kijima et al., 1996). This personality test built a model for temperament with a physiological basis in the background. The test consists of 240 items using a yes-no scale rating.
Cloninger hypothesized that personality consists of traits that are hereditary and stable throughout life, and traits mature throughout life under the influence of socio-cultural environment. The TCI consists of seven dimensions, including four temperament dimensions (Novelty Seeking, Harm Avoidance, Reward Dependence, and Persistence) and three character dimensions (Self-directedness, Cooperativeness and Self-transcendence). Three of the temperament dimensions have been hypothesized to be associated with monoamine neurotransmitters. Novelty seeking has been hypothesized to be associated with dopaminergic, harm avoidance with serotonergic, and reward dependence with noradrenergic.

Overview of AI Technology

Current AI technology often uses a technique called machine learning. The technology, such as deep learning, is based on a neural network that consists of perceptrons, which simulate the mechanisms of human neurons in a multilayered network. This makes it feasible to learn various characteristics contained in the data with dramatically high accuracy, in comparison with the conventional techniques (LeCun, Bengio & Hilton, 2015). By inputting a large amount of image, voice, and text data into the neural network, the computer system automatically learns the characteristics contained in these data. There are various alternatives other than deep learning, though AI nowadays often refers to the system that uses it.

Objectives of This Study

In this study, therefore, we examined validity and reliability of blood type personality theory using the previously mentioned academic findings and commonly used methodologies in psychology (e.g. ANOVA). In addition, we used the latest artificial intelligence, AI (machine learning), which can theoretically handle nonlinear multi-factor models. Details are outlined in the following methods section.