The FairCeptron: A Framework for Measuring Human Perceptions of Algorithmic Fairness

Measures of algorithmic fairness often do not account for human perceptions of fairness that can substantially vary between different sociodemographics and stakeholders. The FairCeptron framework is an approach for studying perceptions of fairness in algorithmic decision making such as in ranking or classification. It supports (i) studying human perceptions of fairness and (ii) comparing these human perceptions with measures of algorithmic fairness. The framework includes fairness scenario generation, fairness perception elicitation and fairness perception analysis. We demonstrate the FairCeptron framework by applying it to a hypothetical university admission context where we collect human perceptions of fairness in the presence of minorities. An implementation of the FairCeptron framework is openly available, and it can easily be adapted to study perceptions of algorithmic fairness in other application contexts. We hope our work paves the way towards elevating the role of studies of human fairness perceptions in the process of designing algorithmic decision making systems.

Quota-based debiasing can decrease representation of already underrepresented groups

Many important decisions in societies such as school admissions, hiring, or elections are based on the selection of top-ranking individuals from a larger pool of candidates. This process is often subject to biases, which typically manifest as an under-representation of certain groups among the selected or accepted individuals. The most common approach to this issue is debiasing, for example via the introduction of quotas that ensure proportional representation of groups with respect to a certain, often binary attribute. Cases include quotas for women on corporate boards or ethnic quotas in elections. This, however, has the potential to induce changes in representation with respect to other attributes. For the case of two correlated binary attributes we show that quota-based debiasing based on a single attribute can worsen the representation of already underrepresented groups and decrease overall fairness of selection. We use several data sets from a broad range of domains from recidivism risk assessments to scientific citations to assess this effect in real-world settings. Our results demonstrate the importance of including all relevant attributes in debiasing procedures and that more efforts need to be put into eliminating the root causes of inequalities as purely numerical solutions such as quota-based debiasing might lead to unintended consequences.

The Effects of Randomness on the Stability of Node Embeddings

We systematically evaluate the (in-)stability of state-of-the-art node embedding algorithms due to randomness, i.e., the random variation of their outcomes given identical algorithms and graphs. We apply five node embeddings algorithms---HOPE, LINE, node2vec, SDNE, and GraphSAGE---to synthetic and empirical graphs and assess their stability under randomness with respect to (i) the geometry of embedding spaces as well as (ii) their performance in downstream tasks. We find significant instabilities in the geometry of embedding spaces independent of the centrality of a node. In the evaluation of downstream tasks, we find that the accuracy of node classification seems to be unaffected by random seeding while the actual classification of nodes can vary significantly. This suggests that instability effects need to be taken into account when working with node embeddings. Our work is relevant for researchers and engineers interested in the effectiveness, reliability, and reproducibility of node embedding approaches.

Word-Emoji Embeddings from large scale Messaging Data reflect real-world Semantic Associations of Expressive Icons

We train word-emoji embeddings on large scale messaging data obtained from the Jodel online social network. Our data set contains more than 40 million sentences, of which 11 million sentences are annotated with a subset of the Unicode 13.0 standard Emoji list. We explore semantic emoji associations contained in this embedding by analyzing associations between emojis, between emojis and text, and between text and emojis. Our investigations demonstrate anecdotally that word-emoji embeddings trained on large scale messaging data can reflect real-world semantic associations. To enable further research we release the Jodel Emoji Embedding Dataset (JEED1488) containing 1488 emojis and their embeddings along 300 dimensions.

Joint Multiclass Debiasing of Word Embeddings

Bias in Word Embeddings has been a subject of recent interest, along with efforts for its reduction. Current approaches show promising progress towards debiasing single bias dimensions such as gender or race. In this paper, we present a joint multiclass debiasing approach that is capable of debiasing multiple bias dimensions simultaneously. In that direction, we present two approaches, HardWEAT and SoftWEAT, that aim to reduce biases by minimizing the scores of the Word Embeddings Association Test (WEAT). We demonstrate the viability of our methods by debiasing Word Embeddings on three classes of biases (religion, gender and race) in three different publicly available word embeddings and show that our concepts can both reduce or even completely eliminate bias, while maintaining meaningful relationships between vectors in word embeddings. Our work strengthens the foundation for more unbiased neural representations of textual data.

The POLAR Framework: Polar Opposites Enable Interpretability of Pre-Trained Word Embeddings

We introduce POLAR - a framework that adds interpretability to pre-trained word embeddings via the adoption of semantic differentials. Semantic differentials are a psychometric construct for measuring the semantics of a word by analysing its position on a scale between two polar opposites (e.g., cold -- hot, soft -- hard). The core idea of our approach is to transform existing, pre-trained word embeddings via semantic differentials to a new "polar" space with interpretable dimensions defined by such polar opposites. Our framework also allows for selecting the most discriminative dimensions from a set of polar dimensions provided by an oracle, i.e., an external source. We demonstrate the effectiveness of our framework by deploying it to various downstream tasks, in which our interpretable word embeddings achieve a performance that is comparable to the original word embeddings. We also show that the interpretable dimensions selected by our framework align with human judgement. Together, these results demonstrate that interpretability can be added to word embeddings without compromising performance. Our work is relevant for researchers and engineers interested in interpreting pre-trained word embeddings.

Privacy Attacks on Network Embeddings

Data ownership and data protection are increasingly important topics with ethical and legal implications, e.g., with the right to erasure established in the European General Data Protection Regulation (GDPR). In this light, we investigate network embeddings, i.e., the representation of network nodes as low-dimensional vectors. We consider a typical social network scenario with nodes representing users and edges relationships between them. We assume that a network embedding of the nodes has been trained. After that, a user demands the removal of his data, requiring the full deletion of the corresponding network information, in particular the corresponding node and incident edges. In that setting, we analyze whether after the removal of the node from the network and the deletion of the vector representation of the respective node in the embedding significant information about the link structure of the removed node is still encoded in the embedding vectors of the remaining nodes. This would require a (potentially computationally expensive) retraining of the embedding. For that purpose, we deploy an attack that leverages information from the remaining network and embedding to recover information about the neighbors of the removed node. The attack is based on (i) measuring distance changes in network embeddings and (ii) a machine learning classifier that is trained on networks that are constructed by removing additional nodes. Our experiments demonstrate that substantial information about the edges of a removed node/user can be retrieved across many different datasets. This implies that to fully protect the privacy of users, node deletion requires complete retraining - or at least a significant modification - of original network embeddings. Our results suggest that deleting the corresponding vector representation from network embeddings alone is not sufficient from a privacy perspective.

iLCM - A Virtual Research Infrastructure for Large-Scale Qualitative Data

The iLCM project pursues the development of an integrated research environment for the analysis of structured and unstructured data in a "Software as a Service" architecture (SaaS). The research environment addresses requirements for the quantitative evaluation of large amounts of qualitative data with text mining methods as well as requirements for the reproducibility of data-driven research designs in the social sciences. For this, the iLCM research environment comprises two central components. First, the Leipzig Corpus Miner (LCM), a decentralized SaaS application for the analysis of large amounts of news texts developed in a previous Digital Humanities project. Second, the text mining tools implemented in the LCM are extended by an "Open Research Computing" (ORC) environment for executable script documents, so-called "notebooks". This novel integration allows to combine generic, high-performance methods to process large amounts of unstructured text data and with individual program scripts to address specific research requirements in computational social science and digital humanities.

How Users Explore Ontologies on the Web: A Study of NCBO's BioPortal Usage Logs

Ontologies in the biomedical domain are numerous, highly specialized and very expensive to develop. Thus, a crucial prerequisite for ontology adoption and reuse is effective support for exploring and finding existing ontologies. Towards that goal, the National Center for Biomedical Ontology (NCBO) has developed BioPortal---an online repository designed to support users in exploring and finding more than 500 existing biomedical ontologies. In 2016, BioPortal represents one of the largest portals for exploration of semantic biomedical vocabularies and terminologies, which is used by many researchers and practitioners. While usage of this portal is high, we know very little about how exactly users search and explore ontologies and what kind of usage patterns or user groups exist in the first place. Deeper insights into user behavior on such portals can provide valuable information to devise strategies for a better support of users in exploring and finding existing ontologies, and thereby enable better ontology reuse. To that end, we study and group users according to their browsing behavior on BioPortal using data mining techniques. Additionally, we use the obtained groups to characterize and compare exploration strategies across ontologies. In particular, we were able to identify seven distinct browsing-behavior types, which all make use of different functionality provided by BioPortal. For example, Search Explorers make extensive use of the search functionality while Ontology Tree Explorers mainly rely on the class hierarchy to explore ontologies. Further, we show that specific characteristics of ontologies influence the way users explore and interact with the website. Our results may guide the development of more user-oriented systems for ontology exploration on the Web.

A System for Probabilistic Linking of Thesauri and Classification Systems

This paper presents a system which creates and visualizes probabilistic semantic links between concepts in a thesaurus and classes in a classification system. For creating the links, we build on the Polylingual Labeled Topic Model (PLL-TM). PLL-TM identifies probable thesaurus descriptors for each class in the classification system by using information from the natural language text of documents, their assigned thesaurus descriptors and their designated classes. The links are then presented to users of the system in an interactive visualization, providing them with an automatically generated overview of the relations between the thesaurus and the classification system.