. The first step in allowing the ethical and responsible use of AI and ML is to establish confidence in the
. The necessity of humans-in-the-loop is highlighted in the more technical areas of model creation. The development of responsible AI is influenced significantly by
IBM AI Ethics[3] talks about 5 pillars to be considered for bringing and developing Ethical AI while [4] talks about Explainability, Accuracy, Bias pillars for Ethical AI.\n
\n\nBecause of the inherent human engagement and context applied to the AI development process, ethical considerations are made at each level of the model creation process to ensure the
and final products. The challenges being faced while applying AI Ethics in practice are also addressed by some authors as the pillars of AI Ethics. Thus following we are briefing these 5 challenges in detail.\n\n
\n \nOften problems occur during the earlier stages of conceptualization, research, and design of AI models. Understanding the rationale behind an AI algorithm’s results (explainability) and the algorithm’s accuracy and consistency (robustness) in producing results contributes to building responsible models.\n\nDiscrepancies may arise at any step in the training data, model, or manufacturing process. You can enhance your models if you understand how and why they do what they do. Furthermore, promptly updating and correcting differences may lead to a more ethical, responsible, and simplified process for future builds and products.\n\nAssuming all AI technology firms continue to value explainability and robustness as concepts. In such situation, it may continue to increase public trust in AI as a relevant and safe technology to develop.\n \n
is a necessary step in machine learning. Machines assimilate on what we tell them. However, the way we define data is subjective. This is when biases such as
used to develop the tools, and it is more common among disadvantaged or marginalized populations, categorizations, or data sets.\n\n
1. mistakes that are naturally woven into the machine learning process based on previous failures
\n 2. human-based errors in human data collecting.
\n\nTeams must be aware of accurate, equitable, and high-quality data representation – specifically, whether data in the sample over- or under-represents a given group within the data collection. To ensure the integrity and correctness of the data, data must be consistently represented in the same manner across the dataset.\n\n 3. Robustness
\n\nThe goal of machine learning training and testing is to assess and improve your model’s accuracy.
\n \n 3.1 Quality of Data
\n \n A well-documented development and testing process can help mitigate and correct unintended risks from inaccurate predictions. If errors occur, having highly accurate training data can help assess how likely the errors are.\n\nMachine learning enables machines to learn independently, but a machine can’t distinguish between good and bad data. Machines must learn from a training dataset or input data, and the quality of that data is directly proportional to a machine learning model’s success.\n\nSo, if a machine learning model delivers inaccurate or irrelevant results, the first step in troubleshooting is evaluating data quality.\n\nSubpar data quality is a critical problem that engineers can’t overlook. Poor data quality can create data cascades that cause adverse downstream effects. For example, these errors can work their way down the development process. These can jeopardize the quality of the decision-making process around refinement and design.\n\nThis is why data quality is vital to ML performance and the value it provides to the operation where it’s deployed.
\n \n\n 3.2 Quantity of data
\n\n \n Insufficient data sets may not accurately reflect all of the qualities that need to be accounted for in a model to produce a consistent outcome. Data must be as complete as possible before moving to the next part of the production process. Ideally, datasets should represent all aspects of operating conditions in a balanced way.\n\n 4. Transparency:
\n \nAny feature of a model should be scrutinized. Is it feasible to understand the model's background? Who built it, what data was utilized, what algorithms, what programs were used, who approved it, and who validated it? All of these model traits and information should be available. Like when you buy food, there is a label on them that gives the nutritional information, when and where it was created, and other information. Similarly, you should be able to quickly discover the model's details.\n\n 5. Privacy:
\n\nPrivacy[10] is often cited as one of the guiding principles of Ethical AI. This is evocative of the need to apply global privacy and data protection framework's general privacy principles to AI/ML systems that handle personally identifiable information. Limiting data collection, guaranteeing data quality, specifying data usage, setting limits on data reuse, holding data users accountable, and inviting them to participate are all parts of this process. \n \n","metadata":{}},{"cell_type":"code","source":"PIL.Image.open(\"/kaggle/input/images/7 images.png\")\n\n","metadata":{"_kg_hide-input":false,"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# Section 2 Research Impact Assessment
\n\n \n\n","metadata":{}},{"cell_type":"markdown","source":"Data Acquisition Source and Dataset Info \n ","metadata":{}},{"cell_type":"code","source":"PIL.Image.open(\"/kaggle/input/graphs/year frequency.png\")","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"The dataset provided by the Kaggle Competition is comprised of a variety of topic papers belonging to the competition's various categories. After extracting the AI Ethics papers from the provided Kaggle dataset, we discovered that only nine papers pertained to this subject. Given in Annexure 1, the relevant preprocessing of the dataset to derive the AI Ethics papers from the Kaggle dataset is described. Due to the time constraints, we have chosen to create our own dataset using a single academic datasource in order to expand the research's scope and reduce the number of sources used to gather papers. This will help us identify the current research trend in AI ethics. Therefore, we have chosen the most popular Academic Database, Web of Science, to compile AI Ethics research papers. The search term (\"AI Ethics\") was used to compile the relevant papers for our topic, and we searched only the titles section of the papers. Thus, using the aforementioned search query, we were able to obtain sixty AI Ethics papers, which we used to compile our dataset for analysis. All research evaluations performed in this segment utilize only these 60 articles from our dataset.
\n","metadata":{}},{"cell_type":"markdown","source":"📌 Import libraries/module.
\n","metadata":{"_kg_hide-input":true,"_kg_hide-output":true}},{"cell_type":"code","source":"#importing libraries\nimport pandas as pd\nimport numpy as np\nimport scipy.stats as st\nimport os\nimport matplotlib.pyplot as plt\nimport seaborn as sns","metadata":{"_kg_hide-output":true,"_kg_hide-input":true,"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"
📌 reading writeup dataset.
\n","metadata":{"_kg_hide-input":true}},{"cell_type":"code","source":"#reading writeup dataset\ndataset_writeup = pd.read_csv(\"/kaggle/input/2023-kaggle-ai-report/kaggle_writeups_20230510.csv\")","metadata":{"_kg_hide-input":true,"_kg_hide-output":true,"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"
📌 understanding Data.
\n","metadata":{"_kg_hide-input":true}},{"cell_type":"code","source":"\ndataset_writeup\n","metadata":{"_kg_hide-input":true,"_kg_hide-output":true,"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"","metadata":{}},{"cell_type":"markdown","source":"
📌 As we can see this dataset belongs to kaggle competition and it is out of our domain respectively .
\n \n","metadata":{"_kg_hide-input":true}},{"cell_type":"markdown","source":"
📌 importing the json data file .
\n \n","metadata":{"_kg_hide-input":true}},{"cell_type":"code","source":"def reading_file():\n with open('/kaggle/input/2023-kaggle-ai-report/arxiv_metadata_20230510.json', 'r') as f:\n for line in f:\n yield line","metadata":{"_kg_hide-input":true,"_kg_hide-output":true,"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"
📌 Reading the contents from dataset .
\n \n","metadata":{"_kg_hide-input":true}},{"cell_type":"code","source":"import json\nread=reading_file()\nfor dataset_read in read:\n for k, v in json.loads(dataset_read).items():\n print(f'{k}: {v}')\n break","metadata":{"_kg_hide-input":true,"_kg_hide-output":true,"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":" 📌 Feature extraction.
\n \n","metadata":{"_kg_hide-input":true}},{"cell_type":"code","source":"titles, abstracts, versions, categories, doi, authors_parsed = [], [], [], [], [], []\n\nread = reading_file()\n# loop over all dataset\nfor dataset_read in read:\n # extract single dataset\n paper_dict = json.loads(dataset_read)\n version = paper_dict.get('versions')\n try:\n versions.append(version[-1]['created']) # choose created as the most recent version\n titles.append(paper_dict.get('title'))\n abstracts.append(paper_dict.get('abstract'))\n categories.append(paper_dict.get('categories'))\n doi.append(paper_dict.get('doi'))\n authors_parsed.append(paper_dict.get('authors_parsed'))\n except:\n pass","metadata":{"_kg_hide-input":true,"_kg_hide-output":true,"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"
📌 converting json data into dataframe for better analysis
.\n \n","metadata":{"_kg_hide-input":true}},{"cell_type":"code","source":"dataset_read = pd.DataFrame({\n 'titles': titles,\n 'abstracts': abstracts,\n 'categories': categories,\n 'versions': versions,\n 'doi': doi,\n 'authors_parsed': authors_parsed\n})\ndataset_read.head()\n\n# reduce memory constraints\ndel titles, abstracts, versions, categories, doi, authors_parsed","metadata":{"_kg_hide-input":true,"_kg_hide-output":true,"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"
📌 Reading the first 5 contents from dataset .
\n \n","metadata":{"_kg_hide-input":true}},{"cell_type":"code","source":"dataset_read.head()","metadata":{"_kg_hide-input":true,"_kg_hide-output":true,"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":" 📌 pre-processing the contents from dataset for abstracts and latest version for extracting month and year .
\n \n","metadata":{"_kg_hide-input":true}},{"cell_type":"code","source":"# clean abstracts\ndataset_read['abstracts'] = dataset_read['abstracts'].apply(lambda x: x.replace(\"\\n\",\"\"))\ndataset_read['abstracts'] = dataset_read['abstracts'].apply(lambda x: x.strip())\n\n# extract date time info \ndataset_read['DateTime']= pd.to_datetime(dataset_read['versions'])\ndataset_read['month'] = pd.DatetimeIndex(dataset_read['DateTime']).month\ndataset_read['year'] = pd.DatetimeIndex(dataset_read['DateTime']).year\n\n# clean authors\ndataset_read['authors_parsed']= dataset_read['authors_parsed'].apply(lambda authors_parsed:[(\" \".join(a)).strip() for a in authors_parsed])\ndataset_read.head()","metadata":{"_kg_hide-input":true,"_kg_hide-output":true,"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"
📌 pre-processing the contents from dataset for latest version to extract date.
\n \n","metadata":{"_kg_hide-input":true}},{"cell_type":"code","source":"dataset_read['date'] = pd.DatetimeIndex(dataset_read['DateTime']).date\n\n","metadata":{"_kg_hide-input":true,"_kg_hide-output":true,"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"dataset_read.head()","metadata":{"_kg_hide-input":true,"_kg_hide-output":true,"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"
📌 checking the datset information as the changes.
\n \n","metadata":{"_kg_hide-input":true}},{"cell_type":"code","source":"dataset_read.info()","metadata":{"_kg_hide-input":true,"_kg_hide-output":true,"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"
📌 checking the column indexes.
\n \n","metadata":{"_kg_hide-input":true}},{"cell_type":"code","source":"dataset_read.columns\n","metadata":{"_kg_hide-input":true,"_kg_hide-output":true,"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"
📌 checking the data type for each column.
\n \n","metadata":{"_kg_hide-input":true}},{"cell_type":"code","source":"dataset_read.dtypes\n","metadata":{"_kg_hide-input":true,"_kg_hide-output":true,"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":" 📌
data cleaning and checking duplicacy of data is present or not.
\n \n","metadata":{"_kg_hide-input":true}},{"cell_type":"code","source":"dataset_read.nunique\n","metadata":{"_kg_hide-input":true,"_kg_hide-output":true,"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"
📌 extracting all the publication happened per year.
\n \n","metadata":{"_kg_hide-input":true}},{"cell_type":"code","source":"dataset_read['year'].value_counts()","metadata":{"_kg_hide-input":true,"_kg_hide-output":true,"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"
📌 plotting the graphs per year.
\n \n","metadata":{"_kg_hide-input":true}},{"cell_type":"code","source":"plt.figure(figsize=(20,5));\nsns.barplot(x=dataset_read['year'].value_counts().head(25).index,palette='winter_r',y=dataset_read['year'].value_counts().head(25).values)\nplt.xticks(rotation = 45)\nplt.xlabel(' Year')\nplt.ylabel('Number of publication each year')\nplt.title('publication per years')","metadata":{"_kg_hide-input":true,"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"
📌 attribution selection for better analysis for categories and formatting abstracts
\n \n","metadata":{"_kg_hide-input":true}},{"cell_type":"code","source":"dataset_read['num_categories'] = dataset_read['categories'].apply(lambda x:len(x)).astype('int')\ndataset_read['num_words_abstract'] = dataset_read['abstracts'].apply(lambda x:len(x.split())).astype('int')\ndataset_read['num_authors'] = dataset_read['authors_parsed'].apply(lambda x:len(x)).astype('int')\ndataset_read.head()","metadata":{"_kg_hide-input":true,"_kg_hide-output":true,"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"
📌 Assuming AI Ethics may come under this categories.
\n \n","metadata":{"_kg_hide-input":true}},{"cell_type":"code","source":"ai_list=['cs.AI','cs.LG','stat.ML','cs.CY']\ndataset_read['is_ai_ethics'] = dataset_read['categories'].apply((lambda x: any(ele in x for ele in ai_list)==True))\ndataset_read = dataset_read[dataset_read['is_ai_ethics']==True]\ndataset_read.head","metadata":{"_kg_hide-input":true,"_kg_hide-output":true,"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"
📌dropping all the un-necessaries rows as per the problem statement.
\n \n","metadata":{"_kg_hide-input":true}},{"cell_type":"code","source":"index_year = dataset_read[ (dataset_read['year'] <=2021)].index\ndataset_read.drop(index_year , inplace=True)\ndataset_read.head(15)","metadata":{"_kg_hide-input":true,"_kg_hide-output":true,"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"dataset_read.info()","metadata":{"_kg_hide-input":true,"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"
📌Checking the available data after deleting the data.
\n \n","metadata":{"_kg_hide-input":true}},{"cell_type":"markdown","source":" 📌searching ethics word in titles present in the paper.
\n \n","metadata":{"_kg_hide-input":true}},{"cell_type":"code","source":"\ndataset_titles = dataset_read[dataset_read['titles'].str.contains('ethics')]\ndataset_titles","metadata":{"_kg_hide-input":true,"_kg_hide-output":true,"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"
📌searching \"ethics\" word in abstract present in the paper.
\n","metadata":{"_kg_hide-input":true}},{"cell_type":"code","source":"\ndataset_titles = dataset_read[dataset_read['abstracts'].str.contains('ethics')] \ndataset_titles","metadata":{"_kg_hide-input":true,"_kg_hide-output":true,"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"
📌 Please note this Scientometric analysis has been analysed software.
\n","metadata":{"_kg_hide-input":true}},{"cell_type":"markdown","source":"Keyword Analysis \n ","metadata":{}},{"cell_type":"markdown","source":"
📌Keywords Analysis
","metadata":{}},{"cell_type":"code","source":"PIL.Image.open(\"/kaggle/input/images/IMAGES/keywords .png\")","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":" \n\nAccording to the number of documents in our dataset, the keywords have been extracted and presented in the form of a graph to analyze the frequent keywords used in the AI Ethics research papers. As per the visualization, we can see that the term “artificial intelligence” is having a bigger circle followed by the term “ethics” and “ai ethics”.\n
","metadata":{}},{"cell_type":"markdown","source":"
📌 we filtered the top 29 most frequently used keywords
","metadata":{}},{"cell_type":"code","source":"PIL.Image.open(\"/kaggle/input/images/IMAGES/selected keywords 29.png\")","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":" \nIn this visualization you can see that we have removed the keywords that are not relevant to “AI Ethics” such as “Buddishm”,”future”,”text” and we have extracted the keywords and presented in the form of a graphs for better analyzation of frequently used keywords in the field of “AI Ethics”
","metadata":{}},{"cell_type":"markdown","source":" 📌 Frequency of occurence of keywords 29
\n\n","metadata":{}},{"cell_type":"code","source":"import PIL # We will import the packages at \"use-time (just for this kernel)\nPIL.Image.open(\"/kaggle/input/graphs/occurrences vs. keyword 29 .png\")","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":" \nThe above graphical visualization provides the same statistics which have been presented in the below bar graph. The bar graph elucidates the occurrences of the keywords identified in various papers that are published in journals and conference Proceedings by different Authors.\n
","metadata":{}},{"cell_type":"markdown","source":"
📌 we filtered the top 16 most frequently used keywords
","metadata":{}},{"cell_type":"code","source":"PIL.Image.open(\"/kaggle/input/images/IMAGES/selected keywords 14.png\")","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":" \nIn this visualization you can see that we have presented the keywords that are domain specific to “AI Ethics” such as “AI Ethics”,”Ethics Guildlines” and presented in the form of a graphs for better analyze of most frequently used keywords in the field of “AI Ethics”\n
\n","metadata":{}},{"cell_type":"markdown","source":" 📌 Frequency of occurence of keywords 16
\n\n","metadata":{}},{"cell_type":"code","source":"import PIL # We will import the packages at \"use-time (just for this kernel)\nPIL.Image.open(\"/kaggle/input/graphs/occurrences vs. keyword.png\")","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":" \nThe above graphical visualization provides the same statistics which have been presented in the below bar graph. The bar graph elucidates the most frequently occurred keywords identified in various papers that are used by various Authors.
\n","metadata":{}},{"cell_type":"markdown","source":"Top Journals \n ","metadata":{}},{"cell_type":"markdown","source":"
📌 Journals
\n","metadata":{}},{"cell_type":"code","source":"PIL.Image.open(\"/kaggle/input/images/IMAGES/journals sources.png\")","metadata":{"_kg_hide-input":true,"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":" \nAccording to the number of documents in our dataset, the Source Of Journals have been extracted and presented in the form of a graph to analyze the highest number of published papers in the specific journals for AI Ethics research papers. As per the visualization, we can see that the term “Science and Engineering intelligence” is having a bigger circle followed by the term “IEEE Technology and Society ”.\n
","metadata":{}},{"cell_type":"markdown","source":"
📌 Top Journals in AI Ethics
\n\n","metadata":{}},{"cell_type":"code","source":"PIL.Image.open(\"/kaggle/input/images/IMAGES/jornals.png\")","metadata":{"_kg_hide-input":true,"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":" \nIn the subsequent graphs, we have depicted the strength of the links between the sources of journals in a year-wise manner, including other journals as well. Notably, there is a larger cluster with a pink outline on the outer circle, indicating that this cluster represents the most cited journals among all the sources. The pink outline highlights the significance of this cluster in terms of citations, suggesting that these journals have received a higher number of references compared to others.
","metadata":{}},{"cell_type":"markdown","source":"
📌 Frequency of occurence of sources and journals
\n\n","metadata":{}},{"cell_type":"code","source":"PIL.Image.open(\"/kaggle/input/graphs/sources and journals.png\")","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":" \nThe above graphical visualization provides the same statistics which have been presented in the below bar graph. The bar graph elucidates the most prominent journal sources\n
\n","metadata":{}},{"cell_type":"markdown","source":"
📌 Frequency of papers published in each journal
\n\n","metadata":{}},{"cell_type":"code","source":"PIL.Image.open(\"/kaggle/input/graphs/paper count based on publishers.png\")","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":" \nIn addition, we have provided the statistical Representation for occurrence of papers published in each journal on the basis of the publishers.As per the visualization,it indicates that \"Springer\" has made a significant contribution to the conference with a large volume of published papers.\n
","metadata":{}},{"cell_type":"markdown","source":"countries Analysis \n ","metadata":{}},{"cell_type":"markdown","source":"
📌 Countries that worked on AI Ethics
\n\n","metadata":{}},{"cell_type":"code","source":"PIL.Image.open(\"/kaggle/input/images/IMAGES/countries.png\")","metadata":{"_kg_hide-input":true,"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":" \nAccording to the number of documents in our dataset, the countries have been extracted and presented in the form of a graph to analyze the contribution of countries at the forefront of research, advancements and initiatives in the field of “AI Ethics”.As per the visualization, we can see that the country “U.S.A” is having a bigger circle followed by the term “Germany” and “England”.
","metadata":{}},{"cell_type":"markdown","source":"
📌 Top Countries that worked on AI Ethics
\n\n","metadata":{}},{"cell_type":"code","source":"PIL.Image.open(\"/kaggle/input/images/IMAGES/selected countries author.png\")","metadata":{"_kg_hide-input":true,"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":" \nIn this visualization, we have removed those countries “Hungray”,”Japan” who has contributed very less as compared to other prominent countries “U.S.A” that worked together in field of “AI Ethics”.
\n","metadata":{}},{"cell_type":"markdown","source":"
📌 Frequency of occurence of countries
\n\n","metadata":{}},{"cell_type":"code","source":"PIL.Image.open(\"/kaggle/input/graphs/countries.png\")","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":" \nThe above graphical visualization provides the same statistics which have been presented in the below bar graph.The bar graphs represent the occurrence of the countries that are actively engaged in AI Ethics research and initiatives.\n
\n","metadata":{}},{"cell_type":"markdown","source":"Organization Analysis \n ","metadata":{}},{"cell_type":"markdown","source":" 📌 Organizations that worked on AI Ethics
\n\n","metadata":{}},{"cell_type":"code","source":"PIL.Image.open(\"/kaggle/input/images/IMAGES/organizations.png\")","metadata":{"_kg_hide-input":true,"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":" \nAccording to the number of documents in our dataset, the organizations have been extracted and presented in the form of a graph to analyze the initiatives of various organizations in the field of “AI Ethics”.As per the visualization, we can see that the country University Tubingen” is having a bigger circle followed by the term “British LIB” and “Digital Catapult”.\n
\n","metadata":{}},{"cell_type":"markdown","source":" 📌 Top Organizations that worked on AI Ethics
\n\n","metadata":{}},{"cell_type":"code","source":"PIL.Image.open(\"/kaggle/input/images/IMAGES/organiztions 3.png\")","metadata":{"_kg_hide-input":true,"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":" \nIn this visualization, we have removed those organizations“Informational Technology university punjab” who has contributed very less as compared to other prominent organizations “University Tubingen ” that worked together in field of “AI Ethics”.\n
\n","metadata":{}},{"cell_type":"markdown","source":"
📌 Frequency of occurence of Organizations
\n\n","metadata":{}},{"cell_type":"code","source":"PIL.Image.open(\"/kaggle/input/graphs/organizations .png\")","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"The above graphical visualization provides the same statistics which have been presented in the below bar graph.The bar graphs represent the contributions of the Organizations that are actively engaged in AI Ethics research and initiatives.
\n\n","metadata":{}},{"cell_type":"markdown","source":"Authors Analysis\n","metadata":{}},{"cell_type":"markdown","source":"
📌Authors that worked on AI Ethics
\n\n","metadata":{}},{"cell_type":"code","source":"PIL.Image.open(\"/kaggle/input/images/IMAGES/authors.png\")","metadata":{"_kg_hide-input":true,"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"According to the number of documents in our dataset, the Authors have been extracted and presented in the form of a graph to analyze the networthy contributions of authors/ group of authors in the field of “AI Ethics\".
","metadata":{}},{"cell_type":"markdown","source":"
📌Authors that worked on AI Ethics
\n\n","metadata":{}},{"cell_type":"code","source":"PIL.Image.open(\"/kaggle/input/images/IMAGES/Authors_image_justifications.png\")","metadata":{"_kg_hide-input":true,"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":" \nIn this visualization, it has been indicating that contributions of various authors that came together to contribute in the field of “AI Ethics”. It is a part of above graphical visualization graphs as we just explore the smaller circles in above visualization
\n","metadata":{}},{"cell_type":"markdown","source":"Citation Analysis \n ","metadata":{}},{"cell_type":"markdown","source":"
📌 citation Authors that worked on AI Ethics
\n\n","metadata":{}},{"cell_type":"code","source":"PIL.Image.open(\"/kaggle/input/images/IMAGES/citations.png\")","metadata":{"_kg_hide-input":true,"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":" \nAccording to the number of documents in our dataset, the citations has been extracted and presented in the form of a graph to analyze the networthy contributions of research paper that are appreciated and being referenced to find new findings in the field of “AI Ethics”.
\n","metadata":{}},{"cell_type":"markdown","source":"
📌 Top citation Authors that worked on AI Ethics
\n\n","metadata":{}},{"cell_type":"code","source":"PIL.Image.open(\"/kaggle/input/images/IMAGES/selected citations.png\")","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":" \nIn this visualization, we have removed those citation papers who contributions for new findings are very less as compared to other prominent cited papers that has been appreciated in field of “AI Ethics”.
\n\n","metadata":{}},{"cell_type":"markdown","source":" 📌 Frequency of occurence of most citations cited by countries
\n\n","metadata":{}},{"cell_type":"code","source":"PIL.Image.open(\"/kaggle/input/graphs/cited by country.png\")","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":" \nFrom the above graphical visualization we extracted the country of authors to provides the statistics which have been presented in the below bar graph.The bar graphs represent the significant contributions of the citation cited by countries that are actively appreciated their Research work in the field of AI Ethics research and initiatives.
\n\n","metadata":{}},{"cell_type":"markdown","source":" 📌 Frequency of occurence of Top cited Papers
\n\n","metadata":{}},{"cell_type":"code","source":"PIL.Image.open(\"/kaggle/input/graphs/top citation .png\")","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":" \nThe above graphical visualization provides the same statistics which have been presented in the below bar graph.The bar graphs represent the significant contributions of the citation that are actively appreciated their Research work in the field of AI Ethics research and initiatives.\n
\n\n","metadata":{}},{"cell_type":"code","source":"# PIL.Image.open(\"/kaggle/input/images/IMAGES/ai ethics reference 2.png\")\n","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"\n# Section 3 Advancement of AI Ethics Table of Contents
","metadata":{}},{"cell_type":"markdown","source":" Despite the advancements in artificial intelligence, the concerns about the ethics and responsible AI were not given much focus. The terms “Ethical AI”, “AI Ethics”, “Responsible AI” are currently gaining much popularity from the past one decade. Due to evolving technology and the high-rate technology adoption has given rise to think on this perspective of the artificial intelligence for the benefit of the human society and the socio-economic concerns. There have been many incidents such as COMPAS[7], Amazon “Rekognition”[8], Deepfakes[9] that compel us to reconsider the development of artificial intelligence products in more sophisticated and ethically responsible ways.This section talks about the initiatives taken at the different levels of the society (i.e. academic, government, industrial, etc.) to develop artificial intelligence in an ethical and more responsible way.
\n","metadata":{}},{"cell_type":"markdown","source":" Initiatives taken by at the intergovernmental level\n:
\n \n1. A High-Level Expert Group on Artificial Intelligence exists within the European Commission. This released its \"Ethics Guidelines for Trustworthy Artificial Intelligence\"[11] on April 8, 2019. A white paper on quality and trust in artificial intelligence innovation[12] was released on February 19, 2020, by the European Commission's Robotics and Artificial Intelligence Innovation and quality section.The Artificial Intelligence Act[13] was additionally recommended by the European Commission.
\n \n2. An OECD AI Policy Observatory[14] was created by the OECD.
\n \n3. The Recommendation on the Ethics of Artificial Intelligence, the first international standard on AI ethics, was adopted by UNESCO[15] in 2021.
\n\n","metadata":{}},{"cell_type":"markdown","source":" Initiatives taken by various government level\n:\n \n1. The
Obama administration created a Roadmap for AI Policy[16] in the US.Two notable white papers on the
implications and future of AI were published by the Obama administration. The National Institute of Standards and Technology
(NIST) was directed by the
White House to start work on Federal Engagement of AI Standards[17] in 2019 in an executive letter known as the
\"American AI Initiative\" (February 2019).\n\n2. A draft executive order on
\"Guidance for Regulation of Artificial Intelligence Applications\" (\"OMB AI Memorandum\"), published by the Office of Management and Budget
(OMB) in January 2020, was made public by the Trump Administration in the United States. The directive places a focus on the need to fund AI applications, increase public confidence in AI, lower obstacles to using AI, and maintain American AI technology's competitiveness in a global market. Although the necessity for
privacy concerns is acknowledged, there is no additional information on enforcement. The emphasis and priority seem to be on
American AI technological advancements.
Federal organizations are even urged to employ the order to avoid any state laws and rules that a market may deem to be too onerous to follow[18].\n \n3. A
20-Year Community Roadmap for Artificial Intelligence Research in the US[19], published as a draft study by the
Computing Community Consortium (CCC), provided input.\n\n4.
US officials are counseled on the security implications of emerging technology, such as AI, by the
Center for Security and Emerging Technology.\n \n\n \n5. The
OECD AI Principles[22], the Council of Europe's Report
\"Towards Regulation of AI Systems[21],\" China's ethical guidelines for using AI, the
UNESCO [20] Recommendation on the Ethics of AI [23], the High-Level Expert Group on AI established by the
European Commission, and the
UNESCO Recommendation on the Ethics of AI are some notable examples of responsible AI frameworks by public organizations.\n \n6. The release of the
Model AI Governance Framework[24] by Singapore's Personal Data Protection Commission, the extensive work on an AI Auditing Framework[25] by the
U.K. Information Commissioner's Office, and the publication of a Guidance on the Ethical Development and Use of AI by
the Office of The Privacy Commissioner for Personal Data of Hong Kong[26] all highlight the significant role that
privacy regulators play in AI governance.\n","metadata":{}},{"cell_type":"markdown","source":"
Initiatives taken at the academic level\n:\n \n1. The
University of Oxford has three research institutions with a primary interest in AI ethics. The Future of
Humanity Institute has two main areas of interest:
AI Governance[28] and AI Safety[27].John Tasioulas is the director of the
Institute for Ethics in AI, whose main objective is to advance AI ethics as a legitimate topic in contrast to related applied ethics fields.
The Oxford Internet Institute, led by
Luciano Floridi, focuses on the moral implications of current ICT and AI developments.\n\n\n2.
Joanna Bryson co-founded the Centre for Digital Governance at the
Hertie School in Berlin to do study on moral and technological issues.\n\n \n3.
The societal effects of artificial intelligence are being researched at the
NYU,[29] AI Now Institute. The topics of its
multidisciplinary study are bias and inclusion, work and automation, rights and freedoms, and safety and civic infrastructure.\n\n4. The consequences of AI on policy and unemployment are studied by the
Institute for Ethics and Emerging Technologies (IEET)[30][31].\n\n5. The
Technical University of Munich's Institute for Ethics in Artificial Intelligence (IEAI)[32], led by
Christoph Lütge, carries out research in a number of areas, including mobility, employment, healthcare, and sustainability.\n\n6. In order to create a future generation of computer scientists with a worldview that considers the social impact of their work[33],
Barbara J. Grosz, the Higgins Professor of Natural Sciences at the
Harvard. John A. Paulson School of Engineering and Applied Sciences, started the
Embedded Ethics initiative.\n","metadata":{}},{"cell_type":"markdown","source":"
Initiatives taken at the industrial level:\n \n\n\nBeyond that, there are a seemingly endless number of attempts taken by businesses to self-regulate. Additionally, in order to improve the responsible use of artificial intelligence (AI), the business sector has joined forces with the
academic community and nonprofit organizations, such as in the Partnership on AI[34] and the Global Partnership for AI[35]. Additionally, assistance may be obtained through
standardization agencies such as ISO/IEC[36], IEEE[37], and NIST[38].\nCompanies such as
Datagen[39], Galileo[40], and Snorkel AI[1] that operate in the field of Data for AI category offer particular services to safeguard data privacy,
identify data bias early, or give alternate ways for data collecting / creation in order to minimize bias amplification later on in the machine learning lifecycle. These services may be found in the Data for AI category.\n","metadata":{}},{"cell_type":"code","source":"PIL.Image.open(\"/kaggle/input/images/images industrial.png\")","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"image source: https://opendatascience.com/wp-content/uploads/2022/05/eth2.png\n","metadata":{}},{"cell_type":"markdown","source":"#
Section 4 Conclusion
\n
","metadata":{}},{"cell_type":"markdown","source":"
Explainability, Bias, Robustness, Transparency, and Privacy are the five most essential pillars of AI Ethics in practice, and they are discussed in detail in this report. During the evaluation, we determined that the work of \"Anat Elhalal\" on AI Ethics has been cited 449 times. The leading journal, with Springer as the leading publisher of AI ethics articles. In terms of national contributions, the United States practices the most AI Ethics work, followed by Germany and the United Kingdom. Numerous government, academic, and business initiatives have been undertaken to introduce ethics and responsibility to artificial intelligence technology in order to increase its accountability and credibility. Despite the fact that many national and international standards are being developed for AI ethics, putting them into practice is one of the most difficult aspects, which raises questions about the security of the technology and how to prevent malicious activities to avoid data poisoning and algorithm poisoning if the pipeline and the development of AI models and data are made public. Thus, this is the subject of the debate, which has a long way to go before every government, industrial and academic body reaches the specific standards and legal norms for the same.\n
","metadata":{}},{"cell_type":"markdown","source":"#
Reference
\n
","metadata":{"_kg_hide-input":false}},{"cell_type":"markdown","source":"
\n \n 1. https://snorkel.ai/
\n \n 2. Hagendorff, T. The Ethics of AI Ethics: An Evaluation of Guidelines. Minds & Machines 30, 99–120 (2020). Doi: doi.org/10.1007/s11023-020-09517-8
\n \n 3. https://www.ibm.com/impact/ai-ethics
\n \n 4. Felzmann, Heike et al. “Towards Transparency by Design for Artificial Intelligence.” Science and Engineering Ethics 26.6 (2020): 3333–3361. Web.
\n \n 5. Siau, Keng, and Weiyu Wang. \"Artificial intelligence (AI) ethics: ethics of AI and ethical AI.\" Journal of Database Management (JDM) 31.2 (2020): 74-87.
\n \n 6. Jobin, Anna; Ienca, Marcello; Vayena, Effy (2 September 2020). \"The global landscape of AI ethics guidelines\". Nature. 1 (9): 389–399. arXiv:1906.11668. doi:10.1038/s42256-019-0088-2. S2CID 201827642.
\n \n 7. https://www.nytimes.com/2017/05/01/us/politics/sent-to-prison-by-a-software-programs-secret-algorithms.html \n \n 8. https://edition.cnn.com/2019/06/11/tech/zuckerberg-deepfake/index.html#:~:text=The%20clip%20of%20Zuckberg%20used,CBS%20spokesperson%20told%20CNN%20Business.
\n \n 9. http://proceedings.mlr.press/v81/buolamwini18a/buolamwini18a.pdf
\n \n 10. Lee, Raymond ST, and Raymond ST Lee. \"AI Ethics, Security and Privacy.\" Artificial Intelligence in Daily Life (2020): 369-384.
\n \n 11. Ethics guidelines for trustworthy AI\". Shaping Europe’s digital future – European Commission. European Commission. 2019-04-08. Archived from the original on 2020-02-20. Retrieved 2020-02-20.
\n\n 12. \"White Paper on Artificial Intelligence – a European approach to excellence and trust | Shaping Europe's digital future\". Archived from the original on 2021-03-06. Retrieved 2021-03-18.
\n\n 13. \"Why the world needs a Bill of Rights on AI\". Financial Times. 2021-10-18. Retrieved 2023-03-19.\n
\n 14. \"OECD AI Policy Observatory\". Archived from the original on 2021-03-08. Retrieved 2021-03-18.
\n\n 15. UNESCO member states adopt first global agreement on AI ethics\". Helsinki Times. 2021-11-26. Retrieved 2023-04-26.
\n\n 16. \"The Obama Administration's Roadmap for AI Policy\". Harvard Business Review. 2016-12-21. ISSN 0017-8012. Archived from the original on 2021-01-22. Retrieved 2021-03-16.
\n\n 17. \"Accelerating America's Leadership in Artificial Intelligence – The White House\". “trumpwhitehouse.archives.gov”. Archived from the original on 2021-02-25. Retrieved 2021-03-16.
\n\n 18. \"CCC Offers Draft 20-Year AI Roadmap; Seeks Comments\". HPCwire. 2019-05-14. Archived from the original on 2021-03-18. Retrieved 2019-07-22.
\n\n 19. \"Request Comments on Draft: A 20-Year Community Roadmap for AI Research in the US » CCC Blog\". 13 May 2019. Archived from the original on 2019-05-14. Retrieved 2019-07-22.
\n\n 20. (UNESCO’s Recommendation on the Ethics of AI, China’s): https://unesdoc.unesco.org/ark:/48223/pf0000377897
\n\n 21. (Council of Europe’s Report “Towards Regulation of AI Systems”): https://rm.coe.int/prems-107320-gbr-2018-compli-cahai-couv-texte-a4-bat-web/1680a0c17a
\n\n 22. (Organisation for Economic Co-operation and Development (OCED) AI Principles): https://oecd.ai/en/ai-principles
\n\n 23. (European Commission's Ethics Guidelines for trustworthy AI): https://op.europa.eu/en/publication-detail/-/publication/d3988569-0434-11ea-8c1f-01aa75ed71a1
\n\n 24. (Model AI Governance Framework through Singapore’s Personal Data Protection Commission): https://www.pdpc.gov.sg/-/media/files/pdpc/pdf-files/resource-for-organisation/ai/sgmodelaigovframework2.pdf
\n\n 25. (U.K. Information Commissioner’s Office work on AI Auditing Framework): https://ico.org.uk/for-organisations/uk-gdpr-guidance-and-resources/artificial-intelligence/guidance-on-ai-and-data-protection/
\n\n 26. (Hong Kong’s Ethical Development and Use of AI): https://www.pcpd.org.hk/english/resources_centre/publications/files/guidance_ethical_e.pdf
\n\n 27. Grace, Katja, et al. \"When will AI exceed human performance? Evidence from AI experts.\" Journal of Artificial Intelligence Research 62 (2018): 729-754.
\n\n 28. \"China wants to shape the global future of artificial intelligence\". MIT Technology Review. Archived from the original on 2020-11-20. Retrieved 2020-11-29.
\n\n 29. \"New Artificial Intelligence Research Institute Launches\". 2017-11-20. Archived from the original on 2020-09-18. Retrieved 2021-02-21.
\n\n 30. Hughes, James J.; LaGrandeur, Kevin, eds. (15 March 2017). Surviving the machine age: intelligent technology and the transformation of human work. Cham, Switzerland. ISBN 978-3-319-51165-8. OCLC 976407024. Archived from the original on 18 March 2021. Retrieved 29 November 2020.
\n\n 31. Danaher, John (2019). Automation and utopia: human flourishing in a world without work. Cambridge, Massachusetts. ISBN 978-0-674-24220-3. OCLC 1114334813.
\n\n 32. TUM Institute for Ethics in Artificial Intelligence officially opened\". www.tum.de. Archived from the original on 2020-12-10. Retrieved 2020-11-29.
\n\n 33. Communications, Paul Karoff SEAS (2019-01-25). \"Harvard works to embed ethics in computer science curriculum\". Harvard Gazette. Retrieved 2023-04-06.
\n\n 34. (Partnership on AI): https://partnershiponai.org/
\n\n 35. ( Global Partnership for AI): \n https://www.gpai.ai/projects/data-governance/gpai-data-governance-work-framework-paper.pdf
\n\n 36. (Standardization bodies - ISO/IEC): https://www.iso.org/standard/77608.html
\n\n 37. (Standardization bodies - IEEE SA): https://standards.ieee.org/initiatives/autonomous-intelligence-systems/
\n\n 38. (Standardization bodies - NIST): https://www.nist.gov/news-events/events/2021/10/kicking-nist-ai-risk-management-framework
\n\n 39. https://datagen.tech/
\n\n 40. https://www.rungalileo.io/
","metadata":{}},{"cell_type":"markdown","source":" #
Annexure
\n","metadata":{"_kg_hide-input":false}},{"cell_type":"markdown","source":"
Annexure 1:
\n\n In this researched overview on the topic of \"AI Ethics,\" the section containing the information about the Data Acquisition Source and Dataset Info has been hidden. However, if you would like to examine the details and understand how we obtained only 9 papers from the dataset provided by Kaggle, you can copy and edit this notebook to access the code and its corresponding output. By doing so, you will be able to see the specific steps taken to retrieve the 9 papers and analyze the dataset.
","metadata":{}},{"cell_type":"markdown","source":"
Annexure 2:
\n \nTable 1- Ethical guidelines for AI by country of issuer[6]
","metadata":{}},{"cell_type":"code","source":"#importing libraries\nimport pandas as pd\nimport numpy as np\nimport scipy.stats as st\nimport os\nimport matplotlib.pyplot as plt\nimport seaborn as sns","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"#reading writeup dataset\ncountry_table = pd.read_csv(\"/kaggle/input/ai-ethics-our-dataset/table.csv\")\npd.set_option('display.max_rows', None)\ncountry_table\n ","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"PIL.Image.open(\"/kaggle/input/graphs/Count of Country of issuer (2).png\")","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"PIL.Image.open(\"/kaggle/input/graphs/Count of Country of issuer (1).png\")","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"%%writefile submission.csv\ntype,value\nessay_category,'AI ethics'\nessay_url,'https://www.kaggle.com/ritwikdalmia/unlocking-the-promise-ai-ethics-in-action/'\nfeedback1_url,'https://www.kaggle.com/code/acetonewong/what-genai-looks-like-in-2023/comments#2331060'\nfeedback2_url,'https://www.kaggle.com/code/sohamjangra/image-and-video-data-analysis-using-ai/comments#2340215'\nfeedback3_url,'https://www.kaggle.com/code/kausheeki01/analyzing-text-data-for-customer-insights/comments#2340315'","metadata":{"trusted":true},"execution_count":null,"outputs":[]}]}