@article{behroozi_shirolkar_barik_parnin_2020, title={Debugging Hiring: What Went Right and What Went Wrong in the Technical Interview Process}, DOI={10.1145/3377815.3381372}, abstractNote={The typical hiring pipeline for software engineering occurs over several stages—from phone screening and technical on-site interviews, to offer and negotiation. When these hiring pipelines are “leaky,” otherwise qualified candidates are lost at some stage of the pipeline. These leaky pipelines impact companies in several ways, including hindering a company’s ability to recruit competitive candidates and build diverse software teams.To understand where candidates become disengaged in the hiring pipeline—and what companies can do to prevent it—we conducted a qualitative study on over 10,000 reviews on 19 companies from Glassdoor, a website where candidates can leave reviews about their hiring process experiences. We identified several poor practices which prematurely sabotage the hiring process—for example, not adequately communicating hiring criteria, conducting interviews with inexperienced interviewers, and ghosting candidates. Our findings provide a set of guidelines to help companies improve their hiring pipeline practices—such as being deliberate about phrasing and language during initial contact with the candidate, providing candidates with constructive feedback after their interviews, and bringing salary transparency and long-term career discussions into offers and negotiations. Operationalizing these guidelines helps make the hiring pipeline more transparent, fair, and inclusive.}, journal={2020 IEEE/ACM 42ND INTERNATIONAL CONFERENCE ON SOFTWARE ENGINEERING: SOFTWARE ENGINEERING IN SOCIETY (ICSE-SEIS 2021)}, author={Behroozi, Mahnaz and Shirolkar, Shivani and Barik, Titus and Parnin, Chris}, year={2020}, pages={71–80} }
 @article{ford_behroozi_serebrenik_parnin_2019, title={Beyond the Code Itself: How Programmers Really Look at Pull Requests}, DOI={10.1109/ICSE-SEIS.2019.00014}, abstractNote={Developers in open source projects must make decisions on contributions from other community members, such as whether or not to accept a pull request. However, secondary factors-beyond the code itself-can influence those decisions. For example, signals from GitHub profiles, such as a number of followers, activity, names, or gender can also be considered when developers make decisions. In this paper, we examine how developers use these signals (or not) when making decisions about code contributions. To evaluate this question, we evaluate how signals related to perceived gender identity and code quality influenced decisions on accepting pull requests. Unlike previous work, we analyze this decision process with data collected from an eye-tracker. We analyzed differences in what signals developers said are important for themselves versus what signals they actually used to make decisions about others. We found that after the code snippet (x=57%), the second place programmers spent their time fixating is on supplemental technical signals (x=32%), such as previous contributions and popular repositories. Diverging from what participants reported themselves, we also found that programmers fixated on social signals more than recalled.}, journal={2019 IEEE/ACM 41ST INTERNATIONAL CONFERENCE ON SOFTWARE ENGINEERING: SOFTWARE ENGINEERING IN SOCIETY (ICSE-SEIS 2019)}, author={Ford, Denae and Behroozi, Mahnaz and Serebrenik, Alexander and Parnin, Chris}, year={2019}, pages={51–60} }
 @article{behroozi_lui_moore_ford_parnin_2018, title={Dazed: Measuring the Cognitive Load of Solving Technical Interview Problems at the Whiteboard}, url={http://www.scopus.com/inward/record.url?eid=2-s2.0-85049810426&partnerID=MN8TOARS}, DOI={10.1145/3183399.3183415}, abstractNote={Problem-solving on a whiteboard is a popular technical interview technique used in industry. However, several critics have raised concerns that whiteboard interviews can cause excessive stress and cognitive load on candidates, ultimately reinforcing bias in hiring practices. Unfortunately, many sensors used for measuring cognitive state are not robust to movement. In this paper, we describe an approach where we use a head-mounted eye-tracker and computer vision algorithms to collect robust metrics of cognitive state. To demonstrate the feasibility of the approach, we study two proposed interview settings: on the whiteboard and on paper with 11 participants. Our preliminary results suggest that the whiteboard setting pressures candidates into keeping shorter attention lengths and experiencing higher levels of cognitive load compared to solving the same problems on paper. For instance, we observed 60ms shorter fixation durations and 3x more regressions when solving problems on the whiteboard. Finally, we describe a vision for creating a more inclusive technical interview process through future studies of interventions that lower cognitive load and stress.}, journal={2018 IEEE/ACM 40TH INTERNATIONAL CONFERENCE ON SOFTWARE ENGINEERING: NEW IDEAS AND EMERGING TECHNOLOGIES RESULTS (ICSE-NIER)}, author={Behroozi, Mahnaz and Lui, Alison and Moore, Ian and Ford, Denae and Parnin, Chris}, year={2018}, pages={93–96} }
 @article{behroozi_sami_2016, title={A multiple-classifier framework for Parkinson's disease detection based on various vocal tests}, volume={2016}, url={http://www.scopus.com/inward/record.url?eid=2-s2.0-84969752675&partnerID=MN8TOARS}, DOI={10.1155/2016/6837498}, abstractNote={Recently, speech pattern analysis applications in building predictive telediagnosis and telemonitoring models for diagnosing Parkinson’s disease (PD) have attracted many researchers. For this purpose, several datasets of voice samples exist; the UCI dataset named “Parkinson Speech Dataset with Multiple Types of Sound Recordings” has a variety of vocal tests, which include sustained vowels, words, numbers, and short sentences compiled from a set of speaking exercises for healthy and people with Parkinson’s disease (PWP). Some researchers claim that summarizing the multiple recordings of each subject with the central tendency and dispersion metrics is an efficient strategy in building a predictive model for PD. However, they have overlooked the point that a PD patient may show more difficulty in pronouncing certain terms than the other terms. Thus, summarizing the vocal tests may lead into loss of valuable information. In order to address this issue, the classification setting must takewhathas been said into account. As a solution, we introduced a new framework that applies an independent classifier for each vocal test. The final classification result would be a majority vote from all of the classifiers. When our methodology comes with filter-based feature selection, it enhances classification accuracy up to15%.}, journal={International Journal of Telemedicine and Applications}, author={Behroozi, M. and Sami, A.}, year={2016} }
 @inproceedings{behroozi_boostani_2013, title={Presenting a new cascade structure for multiclass problems}, url={http://www.scopus.com/inward/record.url?eid=2-s2.0-84894120576&partnerID=MN8TOARS}, DOI={10.1109/ICECCO.2013.6718261}, abstractNote={Designing a robust and accurate classifier is one of the most important goals in the machine learning society. This issue becomes crucial in the case of multi-class problems. In this research, a new architecture of cascaded classifiers is proposed to handle multi-class tasks. The stages of the proposed cascade are broken into some sub-stages; each contains a number of classifiers. Here, LogitBoost is used as the base classifier due to its low sensitivity to the noisy samples. To assess the proposed method, other cascade structures are implemented and eleven datasets derived from UCI repository are selected as the benchmark. Experimental results imply on the effectiveness of the proposed cascade approach compared to LogitBoost as one of the most successful parallel ensemble structure.}, booktitle={2013 International Conference on Electronics, Computer and Computation, ICECCO 2013}, author={Behroozi, M. and Boostani, R.}, year={2013}, pages={192–195} }