Ahmed Izzidien

Programming artificial intelligence to make fairness assessments of texts through top-down rules, bottom-up training, or hybrid approaches, has presented the challenge of defining cross-cultural fairness. In this paper a simple method is presented which uses vectors to discover if a verb is unfair or fair. It uses already existing relational social ontologies inherent in Word Embeddings and thus requires no training. The plausibility of the approach rests on two premises. That individuals consider fair acts those that they would be willing to accept if done to themselves. Secondly, that such a construal is ontologically reflected in Word Embeddings, by virtue of their ability to reflect the dimensions of such a perception. These dimensions being: responsibility vs. irresponsibility, gain vs. loss, reward vs. sanction, joy vs. pain, all as a single vector. The paper finds it possible to quantify and qualify a verb as fair or unfair by calculating the cosine similarity of the said verb’s embedding vector against FairVec—which represents the above dimensions. We apply this to Glove and Word2Vec embeddings. Testing on a list of verbs produces an F1 score of 95.7, which is improved to 97.0. Lastly, a demonstration of the method’s applicability to sentence measurement is carried out.

We sought to compare the implicit and explicit views of a group of Muslim graduates on the fairness of Islamic law. In this preliminary investigation, we used the Electroencephalographic N400 Event Related Potential to detect the participant’s implicit beliefs. It was found that the majority of participants, eight out of ten, implicitly held that Islamic Law was unfair despite explicitly stating the opposite. In seeking to understand what separated these eight participants from the remaining two – the two who both implicitly and explicitly held that Islamic Law was fair – only two distinguishing characteristics could be identified. Both participants had undertaken an in-depth study of a branch of Islamic law that places the spirit of the law above that of a literal interpretation. They had also attended the same seminary, exclusive to the other participants. Of the eight participants, it was discovered that, while they implicitly held Islamic law to be unfair, they also held it to be rational – in the same way they found that it was rational to push a person of a ship in order to save the remaining from drowning, yet unfair. We discuss these preliminary findings and consider theories on how an innate sense of fairness, an aspect of nativism, may come into play when it is not congruent with a participant’s own beliefs. Further, we ask, where such an inconsistency occurs, how does the mind attempt to rectify it – if at all? As a possible contribution to the discussion on theories of nativism vs. empiricism we put forward a hypothesis and methodology for investigation that may produce previously unconsidered data on human nature.

Programming artificial intelligence to make fairness assessments of texts through top-down rules, bottom-up training, or hybrid approaches, has presented the challenge of defining cross-cultural fairness. In this paper a simple method is presented which uses vectors to discover if a verb is unfair or fair. It uses already existing relational social ontologies inherent in Word Embeddings and thus requires no training. The plausibility of the approach rests on two premises. That individuals consider fair acts those that they would be willing to accept if done to themselves. Secondly, that such a construal is ontologically reflected in Word Embeddings, by virtue of their ability to reflect the dimensions of such a perception. These dimensions being: responsibility vs. irresponsibility, gain vs. loss, reward vs. sanction, joy vs. pain, all as a single vector. The paper finds it possible to quantify and qualify a verb as fair or unfair by calculating the cosine similarity of the said verb’s embedding vector against FairVec—which represents the above dimensions. We apply this to Glove and Word2Vec embeddings. Testing on a list of verbs produces an F1 score of 95.7, which is improved to 97.0. Lastly, a demonstration of the method’s applicability to sentence measurement is carried out.