The level of traction recently gained by the family of generative artificial intelligence tools has amplified calls for an inclusive training set of the language models in order to ensure the tools serves the purpose of less-popular languages as the English language. It is therefore important to ascertain the level of interest in the intelligent tools by societies across the globe. The purpose of this study is to explore the interest of Nigerians in the generative AI tools, as well as the potential socioeconomic factors that may influence their awareness of its use cases. A multivariate metrics containing both socioeconomic and demographic data and as well as Nigeria's web analytics metrics from Google Trends are used for the study. An exploratory data analysis is implemented on the data attributes using python programming to infer actionable insights. Experimental result showed that there is no positive correlation between the literacy level, poverty index, population distribution of Nigerians and their awareness and interest in generative AI tools. The study also revealed that the most popular keywords related to generative AI tools in Nigeria were "Generative AI" and "ChatGPT" even in the northern region with lower literacy level, just as only Lagos returned inquiries on language models.
People all across the world have been fascinated by and interested in artificial intelligence (AI). However, different cultures and geographical areas may have varying level of interest in and involvement with the technology. As a branch of artificial intelligence that deals with the production of original content, such as literature, images, and music, generative AI (gAI) has attracted a lot of attention in recent years. Investigating the degree of interest and involvement in this technology over time across cultures and countries is crucial given its potential effects on creative sectors and the larger society. In recent past, both Language Models (LM) and gAI have gained traction with varying use case perceptions. Indeed, an AI algorithm known as a LM is trained to forecast the likelihood of the subsequent word in a string of words [1]. A LM’s training data often comprise of text from sources like books, papers, and websites. In order to produce convincing and cohesive text, a LM learns the patterns and structures of language from the training data. However, any AI system that can produce new content, such as text, graphics, or music, depending on a collection of input data is referred to as a gAI [2]. One use case of gAI is the LM, but there are other kinds of gAI systems that work with other kinds of input data. A gAI is therefore a broader context of the family of human language-based AI use cases. These AI use cases, especially gAI, have a complex and multifaceted impact on digital literacy across cultures and regions. These technologies provides potent tools for enhancing communication, education, and access to information [3]. Functionalities of chatbots and virtual assistants that can interact with users in natural language, answering questions and offering content-generation assistants, are some of the important utilitarian values. There are also concerns that these technologies may worsen existing inequalities and biases in the global society [4]. A LM trained on a textual data written in English or other languages spoken by more privileged populations may not be as effective at generating text in other languages or dialects that are less well-represented in the training data. Whatever improvement that can scale the representations of less-represented languages in its training set should stem from the popularity of the AI tools in affected world regions. The aim of this study therefore is to analyze and compare the level of awareness and interest in gAI, over time, in different cultures and regions that makes up Nigeria. The adoption of Nigeria serves the tripod purpose of being the most populated country in Africa [5], the largest economy in Africa [6], and the world’s largest black nation [7]. Nigeria’s socioeconomic and demographic data are acquired for this study, alongside its web analytics metrics on gAI. Thereby, the problem statement is to investigate the relationship between demographic, socioeconomic, and digital indicators of Nigeria ethnicities as it relates to their level of gAI awareness and interest. By analyzing these data sets, the observational study would infer insights into how factors such as population, poverty index, literacy level, and digital connectivity metrics relates to Nigerian gAI utility. The rest of the paper is structured in the following ways; section II discusses existing literature while section III explained the EDA methodology. Experimental result is discussed in section IV and the study is concluded with recommendations in section VI.
Artificial intelligence is not new to the global information technology lexicon, as well as its use
cases of machine learning, expert system, natural language processing, fuzzy logic, etc. The sudden
interest in data science is not also completely lost on the less-developed societies across in this
generation of Internet of Things, and an improved discussions on the importance of smart cities. These
tools and their unprecedented functionalities continue to shape global discuss in the academia and
industries. Researchers believe their perceived ease of use and usefulness has helped gain traction in
academia [8]. The entry of Cha
This study involves two phases of data acquisition and the exploratory data analytics. Observational
metrics on Nigeria, referring to quantitative indicators including population, poverty index, literacy
level, and internet subscription, are used in this study. The data attributes of each state of the country
include their Multidimensional Poverty Index (MDI), the Population Size (POP), the Internet
Subscription rate (INT_SUB), their Male gender literacy level (M_LIT_L), the Female gender literacy
level (F_LIT_L), their actual literacy level (LIT_RATE); their search rate for keywords ‘chatgpt’
(CHATGPT), ‘generative AI’ (gAI), ‘artificial intelligence’ (AI), and ‘language model’ (LM). These
data are acquired from the National Bureau of Statistics [17] and the Nigeria Communication
Commission [18], and represent observations within the last one year. The observational data is used
with web analytics data about Nigeria from GT, within the last 12 mon
A dataset's IQR is a measure of variability that sheds light on the distribution of the middle 50% of the data [20]. The IQR is specifically the variation between a dataset's third and first quartiles (Q3 and Q1, respectively). Compared to the range or standard deviation, it is a reliable measure of dispersion that is less susceptible to outliers. The amount of variability inside the middle 50% of the data is one insight that may be obtained from the IQR. A lower IQR indicates that the data are closely grouped around the median, whereas a higher IQR indicates that the data are spread out more widely. The IQR can also be used to spot probable outliers in a dataset, which are often identified as observations that are more than 1.5 times the IQR below Q1 or above Q3 in the dataset.
depicts the most centered value in the 1st half of the rank-organized dataset;
Q1 = { +1}th
4 Q3 = {3 +1}th 4 (1) (2) is the most centered value in the 2nd half of the rank- organized dataset, while Q2 is the median and computed as:
Q2 = Q3 – Q1 (3)
The correlation coefficient is a statistical indicator that displays the strength and direction of the linear relationship between two variables [21]. It is a number between -1 and 1, where a value of 1 implies an idealized absence of correlation between the variables, a value of 1 denotes an idealized presence of correlation, and a value of 0 denotes an idealistic negative correlation. When a variable where x and y are the attributes being investigated for likely positive or negative correlative relationship. The COV is the covariance and is the standard deviation of the two attributes. The heat plot will be employed to display the correlation coefficient matrix of the multivariate data set of this study. Standard deviation and other measures of dispersion including the mean and median will be computed.
3.3 Standard Deviation = √∑( − )2 (4) (5) rises, the other variable also tends to climb if the correlation coefficient is positive. A low correlation value indicates that the other variable tends to decline when the first variable increases. The magnitude of the correlation coefficient, with larger absolute values indicating stronger correlations, determines the strength of the association. In this study, the correlation coefficient will be computed to investigate the linear relationship between the multivariate metrics. The correlation is computed thus:
Correlation = P =
( , )
The rate of deviation of variables contained in the multivariate data from the mean is captured by the value of the standard deviation. Evaluating the variability or dispersion of each variable (such as population, poverty index, literacy level, and internet penetration indices) throughout the Nigerian states with respect to their interest in gAI would be instructive for this study. where is the population standard deviation, N is the population size, xi is each value from the population, and is the the population mean
The eleven (11) attributes represented in the study data, as described earlier in section 3, are for the 36 states and the Federal Capital Territory of Nigeria. The instances describes the socioeconomic and demographic metrics of each of the thirty-seven (37) federating units of the country. The data is further analyzed along the six geo-political zones of the country in order to fully domesticate the findings across Nigeria’s multi-ethnic regions. The statistical summary of the data is presented in Table 1 and Table 2 for the entire multivariate metrics with respect to their measures of dispersion as well as the IQR. Figure 1 shows the rate at which Nigerians enquire about the keywords ‘chatgpt’, ‘generative AI’. ‘language model’, and ‘artificial intelligence’ within the last 12 months on Google search engine. As could be observed, the line plot summed all search per month for the various keywords. The degree of the keyword search per thirty seven federating units is plotted in Figure 2, while Figure 3 shows the plot of Nigeria’s population, multidimensional poverty index, and the literacy level in terms of the geo-political zones. The geo political zone-based plot is likewise presented in Figure 4 for searches on the keywords. The correlation coefficient matrix of the attributes is presented in the heat plot of Figure 5 where the correlation value of each attribute pair is presented across the rows and columns. Line Chart of enquiry on the AI tools
Weekly distribution within the last 12 months
CHATGPT gAI
AI
LM
The statistical summary table is revealing of the spread of data points across critical points. The
multidimensional poverty index of the Nigerian states returns a mean value of 0.2634 for a population
with an average of 5 million people per state. The literacy level of the country within the last one year
is put an average of 50 and their internet subscription average 3.7. With the somewhat high literacy and
internet subscription rate, an average of 81%, 10%, 0.027% and 0.027% enquiries were made on
‘chatgpt’, ‘generative AI’, ‘language model’ and ‘artificial intelligence’ itself respectively. ChatGPT
keyword is the most popular across the country returning a maximum search percentage of 100% per
time with artificial intelligence having a maximum of 28% search enquiry per time. The 25% or less of
total search per time is 80% about chatgpt only. Considering 50% of search history per time, 85% are
about chatgpt while 14% will be about artificial intelligence. Not more than 75% of the entire search
entries are 100% of chatgpt and some 18% of them focused on artificial intelligence keyword. The data
indicates that until the Q2 (median) of the IQR, there are no traces of generative AI, artificial
intelligence, and language model keywords in the enquiry efforts of Nigerian within the last year. As
observed from the statistical summary table, the mean poverty index per a state in Nigeria is 0.264 with
an average population of 5 million people. The mean literacy rate is 50.338, with an average internet
subscription rate of 3 million people per state. Given this, a substantial number of Nigerians has had
internet access in the last one year across each state with over 85% of their search enquiries on ‘chatgpt’,
10% on ‘artificial intelligence’ and an insignificant part of the population on ‘language model’ and
‘generative AI’ keywords. As observed from Figure 1, there was no record of ‘chatgpt’ in the search
history of Nigerians until Nove
The study employed Nigeria’s socioeconomic and demographic data together with web analytics metrics from Google Trend for an exploratory data analysis. The aim is to discover the relationship between the awareness and interest levels on Nigerians on generative artificial intelligence tools and the various factors that could influence the predisposition of different cultures towards the AI tools. Experimental result reveals the popularity of ChatGPT over other keywords like generative AI and language models and as well as the awareness of the intelligent tools even at regions with low literacy and high multidimensional poverty index.
Authors profoundly appreciate the painstaking efforts of the reviewers.