DAFT

Mobile Application and Content Selection

By Amy Solomons

Some context to the upcoming problem…

It has been suggested that low levels of financial literacy cause financial exclusion at the “bottom of the pyramid” in South Africa [1].

Currently, there are Fintech products that endeavor to help people manage their finances.

But, these applications do not have dedicated financial assistants that use Natural Language Generation (NLG) to provide summaries about a user’s budget.

NLG would be useful for cases where non-experts need to interpret complex information [2].

Thus, NLG systems may deem useful in helping the financially illiterate interpret information about their finances and ultimately increase their level of financial literacy efficiently.

However, that's not all. According to a study, only 25% of the population has proficient levels of English [3].

People with low levels of literacy also require assistance in their home languages when visiting banks [3].

So, what is the project problem?

There is a lack of personal finance platforms that aim to reduce financial illiteracy in the South African context by incorporating digital assistants supporting the indigenous languages of South Africa.

Project Goal

Ultimately, the goal of the project is to create a personal finance management mobile application that incorporates NLG to help facilitate the financially illiterate in South Africa. However, in order to obtain this goal 4 objectives were identified:

Incorporate a simplistic solution to reduce financial illiteracy in the application by using a known behaviour model and budgeting technique

Design a suitable application interface 

Identify and use the appropriate content for the NLG feature in the application

Incorporate at least one indigenous language of South Africa in the application

How did I achieve the project goal?

1

An android mobile application was devel​oped on Android Studio. 

2

The application used the Balanced Money formula to reduce financial illiteracy [4]. The application, therefore, contains 3 parent categories called necessitie​s, luxuries and savings to represent the formula. Users are also allowed to create their own sub-categories under the parent categories.

3

The interface was designed on Adobe XD.  Since AI, the parent field of NLG, can create mistrust, inducing trust in the application was required to be considered [8]. The design of the interface was therefore based on Wang and Emurian’s framework on inducing user trust. Their framework states that ease of navigation and colour are important aspects in obtaining user trust in an application [5]. Ease of navigation is created by maintaining consistency and simplicity in the interface, therefore throughout the application constant usage of cards were used to induce familiarity and simplicity for the user [5].  Many of the colours in the interface were also cool-toned colours as suggested in Wang and Emurian’s framework [5]. White and cool shaded greens were used for cards. The most prominent colour in the interface was green as it was determined as one of the colours that induce trust [6, 7].

4

In order to allow the user to store and view information about their budget, a Room da​tabase was implemented. The room database acted as the main storage of information for both the budgeting technique and the content selection. 

5

Unfortunately, due to time constraints the content selection of the NLG feature was brainstormed instead of determined by potential users in initial user interviews. However, the content selection was validated by users during the evaluation process.

6

The application supports isiZulu as it is the first most spoken language outside of South African households [9].  Since the second most spoken language outside of South African households is English, the application also supports English [9].

How did I evaluate my solution?

I evaluated my solution by conducting online questionnaires on LimeSurvey. A non-functional prototype of the application and exemplar output representing the content selected for the NLG feature were evaluated.

The aim of the questionnaires were to determine if the interface was usable, if the design was appropriate, if suitable content for the NLG feature was chosen and if the application was useful.

I planned to recruit 5 English speakers and 5 IsiZulu speakers for the evaluation questionnaires. However, only 5 English speakers and 1 IsiZulu speaker were recurited.

learn more

References

[1] John P Wentzel, Krishna Sundar Diatha and Venkata Seshachal Sarma Yadavalli. 2016. An investigation into factors impacting financial exclusion at the bottom of the pyramid in South Africa. Development Southern Africa 33, 2 (Feb. 2016), 203-214, DOI: https://doi.org/10.1080/0376835X.2015.1120648

[2] Ehud Reiter and Robert Dale. 1997. Building applied natural language generation systems. Nat. Lang. Eng. 3, 1 (March 1997), 57-87. DOI: http://dx.doi.org/10.1017/S1351324997001502

[3]Brigitte Van Schouwenburg. 2009. Taalbeleid aan finansiële instellings. Master’s thesis. University of Johannesburg (UJ), Johannesburg, South africa

[4] Elizabeth Warren and Amelia Warren Tyagi. 2005. All Your Worth: The Ultimate Lifetime Money Plan. Free Press 

[5] Ye Diana Wang and Henry H Emurian. 2005. An overview of online trust: Concepts, elements, and implications. Computers in Human Behavior 21, 1(Jan. 2005), 105–125. DOI: https://doi.org/10.1016/j.chb.2003.11.008 

[6] Ashton Hauff. 2018. The Know It All Guide To Color Psychology In Marketing + The Best Hex Chart. (August 2018). Retrieved on September 09, 2021, from https://coschedule.com/blog/color-psychology-marketing#color-wheel 

[7] Joe Hallock. Color Associations. Retrieved on September 09, 2021, from http://www.joehallock.com/edu/COM498/associations.html

[8] Jungkeun Kim, Marilyn Giroux and Jacob C. Lee. 2021. When do you trust AI? The effect of number presentation detail on consumer trust and acceptance of AI recommendations. Psychology & Marketing 38, 7 (July 2021), 1140–1155. https://doi.org/10.1002/mar.21498 

[9] Saifaddin Galal. 2021. Distribution of languages spoken by individuals inside and outside of households in South Africa 2018. (2021). Retrieved on September 07, 2021, from https://www.statista.com/statistics/1114302/distribution-of-languages-spoken-inside-and-outside-of-households-in-south-africa/#statisticContainer