hahme005

[hahme005]

Final draft

Submitted the last draft on 24/05/2021 and from then I have been proofreading and correcting my dissertation to submit. Since everything has been completed there is nothing else to be added on my trello so this will be the final update of my trello. 

[hahme005]

Fixed report structure

My supervisor in my last meeting told me that the structure of my report was messy and needed to be fixed. I added this to my trello as a task and I have made the structure of my report simpler so that examiners can navigate through it easier. I have now submitted my third draft for review and am working on the 4th and final draft by proofreading the whole report.

[hahme005]

GitHub upload

Uploaded the final version of my survey visualisation in GitHub. I have reduced my survey scope to ask ‘What is the public perception of artificial intelligence?’. I have removed ‘What factors play into a person's perception of artificial intelligence?’ as this question is outside of the scope I would like to discuss. Currently I have 2 research question, the first that will be answered mainly by research and the second will be answered by this survey. I have analysed visualisations and updated my trello to show this! I will now use LaTeX to display this work.

[hahme005]

Second dissertation draft

Submitted my second draft and got some feedback on the structuring of my report and how I present my paragraphs. I have updated my trello!

[hahme005]

First dissertation draft

Submitted my first draft of my dissertation ahead of the meeting I have with my supervisor today. I will keep submitting drafts of my dissertation from now on every 2 weeks. I have updated my trello!

I have also clarified my first research question so it makes more sense, it is now: What characteristics of AI upholds the structure of white supremacy?

[hahme005]

Analysis of visualisations

I have started analysis of the visualisations I will be using for my dissertation. It will be uploaded to my GitHub page. I have not visualised all the questions in my survey and have picked the questions I believe to be the most relevant. Since my topic is an investigation, my chosen questions might change like it did when I finished visualising my data a few months ago. Not all the graphs I have visualised will be used but I will be keeping them there for reference and for if I change my mind. 

My colour scheme is a bright pastels and it will be shown in the GitHub version but the colour range is tailored to the type of question visualised. A diverging colour map is used when the information being plotted has critical middle value. A sequential colour map is used for representing information that has a scale or an ordering. A qualitative colour map is used to represent information that doesn’t have an ordering and to maximise category discrimination. 

Here is an example of the visualisations that uses a qualitative colour map:

The rest will be included in the GitHub page.

[hahme005]

Planning the dissertation report

To complete the bulk of my dissertation I will be splitting the writing portion into small segments written everyday until I reach the word-count. It will then be compiled together and structured accordingly. I have also read the book Artificial Whiteness by Yarden Katz, this book provided relevant insight that I have noted down and will be expanding on. This book gives a view of artificial intelligence and its themes of whiteness and racism in the US. 

I would like to focus mostly on the UK, since there is not a lot of research about white supremacy and AI from a UK perspective. I think I have found a gap in research that I can fill. I would like to use the research question:  Does there exist characteristics in artificial intelligence that upholds the structure of white supremacy?

[hahme005]

Methodology feedback

Received feedback on my methodology on the 23rd March 2021 so I will be applying some changes to accommodate and fix the errors in my work. I will then start working on the bulk of my dissertation and conceptualising the structure of my work.

[hahme005]

Methodology draft 1

Handed in my first methodology draft on 03/03/2021. This included my research questions, methods of survey research, methods of general research with a research proposal. 

I have also completed the visualisations of relevant survey data. These will be included in a section discussing public opinion on artificial intelligence. The secondary dataset, will not be used because it is outside the scope of my project.

Next I will analyse my primary dataset and create a rough structure for my dissertation chapters.

Updated Trello.

[hahme005]

Literature review 1st draft

Submitted my first draft of my literature review on Monday 1st February 2021. I have also created a excel file that includes everything I have read so far, and I have created a review protocol so that my work can be easily recreated for future research. This literature review has made me want to focus of AI racism since research on the topic has been really interesting.

I will be working on my methodology next so that I can start visualising my own survey data. I have updated my Trello to show this.

[hahme005]

Meeting with Supervisor

I had a meeting with my supervisor on Monday to go over my interim report. Below are some important bullet points of things I could have improved on in my report and points that I will use in the future for my final dissertation.

Use more scientific language 

Language was too informal in my interim report

Back up any claim made with references and citations

Cite all images 

Read more scientific papers so that I can replicate the language and structure

Always say ‘to my best knowledge’ when you could not find a thing you are looking for

If you fail at something, note it down and say what you’ve learned from it

Do not remove demographic data, all data is important

Date of meeting: 25th January 2021

[hahme005]

Graphing secondary dataset - exploratory visualisation

After cleaning the dataset, I used Matplotlib, NumPy and Pandas to graph the data. At this stage we are exploring the dataset, the graphs are quick and messy and are used to find pattern that can later be explore in the explanatory stage.

Below are univariate graphs showing the demographic of the survey responders for the secondary dataset.

I have also created multivariate graphs that can be found on my GitHub page

[hahme005]

Plan to create exploratory visualisations for the secondary dataset

Each demographic will have a graph for each question. So there will be three graphs that show the age, education and gender distributions, and then there will be graphs that intersect these demographics with the responses to the questions stated in the previous post. 

Below is a graph that shows those intersections and it will help make sure that all that information has been visualised [created on InVision]

[hahme005]

Secondary dataset

I have yet to find a public dataset related to my project online but a group in my data visualisation module created a survey with a topic similar to mine called ‘AI - A force for good?’. 

I got into contact with one of the group members to ask for the data they collected and thankfully they agreed. I got the excel file of their survey responses, they had 38 responses to their survey and their respondents were university students in the UK.

I have cleaned their dataset using Pandas on Jupyter notebook.

I extracted the questions I was most interested in visualising and ones that were the most similar to my own questions.

Demographic questions they asked:

What is your age?

What is your highest qualification?

Which gender do you most strongly identify with?

Questions similar/related to my survey:

Should the development of Artificial Intelligence be regulated?

Do you fear the potential of Artificial Intelligence?

If you answered 'Yes', what do you think influenced your answer?

Will Artificial Intelligence surpass human intelligence?

Should Artificial Intelligence be able to make moral decisions? - For example: An autonomous car may decide to sacrifice the occupant(s) to avoid hitting a child.

Should Artificial Intelligence be involved in warfare? - Autonomous weapons are currently in development

Do we need Artificial Intelligence for humanity to progress as a species?

Would you be happy to live / work alongside Artificial Intelligence as equals?

Do you believe that Artificial Intelligence will take over jobs that can currently only be performed by humans?

I got these questions from the dataset by using a function in Pandas called usecols. Once I had the columns I needed, I changed the names on each column with a function called series.rename so that the title of each column was easier to work with later on. 

After that I initialised the column age as a category (because the ages are in ranges and a dash is used it cannot be set as a numerical variable), education was initialised as an unordered categorical variable (the reason it was not initialised in the same way as age is because the option ‘prefer not to say’ was removed, I did not think it was a necessary or important addition to the table). 

Here is the cleaned dataset:

Next I will graph this data.

[hahme005]

Creating my dissertation survey

After gathering enough research to create a survey, I started to structure my survey and outline what I wanted to get out of this survey.

First, I needed to make sure I knew what the aim of my survey was: I wanted to find the perception of artificial intelligence held by the public.

Writing down my aim made me question what I meant by ‘public’. It was an extremely broad word to use, and I thought it was a great chance to narrow my scope because a lot of information can be lost when the scope is too broad.

The population for this survey is UK residents.

I have decided to specify my aim to include the population; find the perception of artificial intelligence held by residents in the UK. 

Now that I know my aim and population it was time to create a description for my survey that abided by data protection laws [I needed to explicitly state what I aim to do with this data and that any participant can request the withdrawal of their response]

Next, I needed to produce questions that would help identify the demographic.

The question I am trying to answer to with survey is ‘What factors play into a person’s perception of AI?’.

Factors included in survey:

Age [What is your age?]

In this digital age, how do different generations view the advancement of artificial intelligence

Cite: Wu, J., 2020. How We Educate Our Children Will Change In The Age Of Artificial Intelligence. [online] Forbes. Available at: <https://www.forbes.com/sites/cognitiveworld/2020/02/28/how-we-educate-our-children-will-change-in-the-age-of-artificial-intelligence/?sh=7703ed4d3351> [Accessed 6 January 2021].

Gender [What gender do you identify as?]

Research has shown the existence of gender imbalance in artificial intelligence. ‘AI applications are generally trained using data that are generated by humans, and humans are inherently biased. And many organizations are also biased in their historical behaviour.’

Cite: Wyman, O., n.d. How Artificial Intelligence Perpetuates Gender Imbalance. [online] Oliverwyman.com. Available at: <https://www.oliverwyman.com/our-expertise/insights/2020/mar/gender-bias-in-artificial-intelligence.html> [Accessed 6 January 2021].

Job [What field do you work in?] [What is your experience level in your current job?]

Research has shown that AI has greatly improved the productivity of businesses, mostly in marketing and sales, who does this benefit the most

Cite: Cam, A., Chui, M. and Hall, B., 2019. Global AI Survey: AI Proves Its Worth, But Few Scale Impact. [online] https://www.mckinsey.com/. Available at: <https://www.mckinsey.com/featured-insights/artificial-intelligence/global-ai-survey-ai-proves-its-worth-but-few-scale-impact> [Accessed 6 January 2021].

Race [Which of the following ethnic backgrounds do you identify with?]

Research has shown the failings of artificial intelligence to acknowledge racial prejudice but have these shortcomings been publicised enough to be common knowledge

Research has shown that artificial intelligence can be dangerous to those non-whites. ‘Facial recognition software has trouble identifying the faces of women and black people’. 

Cite: Morris, N., 2020. The Race Problem With AI: ‘Machines Are Learning To Be Racist'. [online] Metro. Available at: <https://metro.co.uk/2020/04/01/race-problem-artificial-intelligence-machines-learning-racist-12478025/> [Accessed 6 January 2021].

This same facial recognition software is used by the metropolitan police to catch criminal activity. Automatic cars hitting black people because it could not recognise black skin.

Household income [What is your current household income?]

Does household income change how you perceive artificial intelligence?

A survey comparing the public perceptions of U.S citizens and senior tech executives showed how 54% of the public believed that AI would hurt the poor while 57% of tech executives disagreed.

Cite: Edelman. 2019. 2019 Artificial Intelligence Survey. [online] Available at: <https://www.edelman.com/research/2019-artificial-intelligence-survey> [Accessed 6 January 2021].

Now that I can identify our demographic the next thing that I needed to do was to present my questions in a structure that made sense.

First, I needed to know if people where coming to this survey with a knowledge of artificial intelligence, what this knowledge consisted of and what influence their view on AI.

Do you have some knowledge of artificial intelligence technology?

What represents artificial intelligence technology the most to you? 

What is your view on the advancement of artificial intelligence technology?

What influenced your view on artificial intelligence the most?

After knowing this I could ask more specific questions on how they viewed the advancement of AI and what technology they would be most comfortable using in the future.

In the near future would you be comfortable with fully autonomous artificial intelligence technology?

What type of artificial intelligence robot are you more comfortable using?

A concept that came up a lot when research perceptions of artificial intelligence was job loss and a fear of un-regulation of this technology. Both these concepts were included.

Do you think that artificial intelligence will surpass human intellect?

Should advancing artificial intelligence technology be regulated?

Do you think artificial intelligence would be able to replace you at your current job?

Some interesting articles I read at this stage talked about the growing distrust of the tech companies that made artificial intelligence software, so I also included question that asked about this, these questions were included to ask if the public feared a loss of control.

How much do you agree with these statements?

Tech companies can be trusted to make ethical artificial intelligence technology

Tech companies are competent in creating safe artificial intelligence technology for the public

Tech companies CAN be trusted to NOT collect our private information from the AI technology we use

Expanding further on the loss of control, questions were asked about how much faith people would put into artificial intelligence.

How much do you agree with these statements?

Privacy will increase as artificial intelligence technology advances

I would trust artificial intelligence technology to make moral decision for me

I believe artificial intelligence cannot discriminate

AI generated news is more accurate and trustworthy than human generated news

The survey finally ends with a question asking people if they believe that the positive impact of the advancement of artificial intelligence would outweigh the negative impact. This question has been asked to solidify the responder and either a pessimist or optimist of the advancement of AI.

How much do you agree with these statements?

The positive impact of the advancement of artificial intelligence greatly outweighs the negative impact

Link to survey: https://forms.gle/qaXmfdhrpVB6A8yP8

[hahme005]

Researching ways to create a good survey

From my data visualisation module this year I have learned the criteria needed to create a good survey that will provide me the answers I need.

Important point to note

Survey introduction/description – needs to be clear and without bias

Wording of questions - needs to be clear, simple, and specific with inclusive language

Relevance – questions need to all be relevant, and it needs to focus on recent past (will provide more accurate answers from responders)

Survey structure – needs to have a logical order to avoid bias

Scope – a good number of questions for sufficient data that can be analysed

Before I could start creating my survey, I needed sufficient information and research that could back up any question I put in my survey. To do this I read through articles and noted down the most interesting points that I could use later.  

[hahme005]

Dissertation project overview

My project will be an investigation into the public perceptions of artificial intelligence. My project aims to find out whether the general public holds a bias against this technology. This project will go through the different ways artificial intelligence technology is used in society by those in varying industries. This will serve to introduce the AI technology I will be discussing for the rest of this project. This project will also discuss the ethics of AI, identifying the main concerns made by the public and evaluating whether these worries are valid or not. It will also question the possibility of another AI winter and if there exists a likelihood of certain fields of AI being underfunded due to certain factors pertaining to public perception.

 What is an AI winter?

From the years 1974–1980 and 1987–1993 was a period where there was a reduced interest and funding in AI research, this event was caused by the disappointment in the aforementioned high expectations of this technology, this was phenomenon was coined the AI winter.

 I plan to carry out surveys that will help me assess the factors that go into moulding a perception of AI from the public. After the surveys, I will visualise and analyse this data with jupyter notebook and python machine learning libraries. This will help in figuring out whether the public perception of AI is positive, neutral, or negative and the circumstances that lead to this conclusion and the effects of this. This is a research project so I will be analysing and interpreting different pieces of work that align with my project to produce the complete scope. My hope is to present a non-bias conclusion to this investigation which will be backed up with first-hand data and research.