Predicting dependency level in dyadic friendship

Abstract

This study aims to (1) investigate the dyadic friendship domain and (2) test it as a predictor of a friendship dependency level. The study was conducted in two stages: Study I and Study 2. Study I explored four friendship domains by using an indigenous psychological approach. Study 2 predicted the dyadic codependency level based on the finding of Study I by implementing an Artificial Neural Network (ANN) with a Multilayer Perceptron (MLP) model conducted in 25 experiments (5 hidden layers x 5 epochs). Data collection was carried out using five open-ended questions and one closed-ended question in Study I; and the Investment Model Scale adaptation questionnaire (13 items; α = 0.832) in Study 2. Study 1 and Study 2 were conducted at a time on 268 respondents (52 male, 216 female). The finding of Study I showed that friendship starts from: (a) the initial contact, which was based on the proximity factor (48.51%), (b) closeness, which was based on personality factors (36.19%), (c) the most frequent activities carried out together in the form of hanging out (52.61%), and (d) the reason for fear of losing a partner due to personality factors (29.48%). Study 2 revealed that friendship dependency level can be predicted by four friendship domains with an accuracy level of 58.35%, in which the initial contact and joint activity domain was of higher importance than the two others (epoch = 5000; hidden layer = 4 units). The overall findings showed that the dyadic codependency level not only can be calculated after friendships are formed and developed but also can be predicted from the initial stages of a relationship when acquaintanceship occurs.

Keywords

• dependensi
• pertemanan
• artificial neural network
• indigenous psychology