How reasonable is it to hold chatbot authors accountable for their bots' behavior?

Community Building Asked by Robert Columbia on November 27, 2020

In a non-Stack Exchange online community, there is a chatbot that runs with the permission of the administrators. The primary purpose of the bot is to analyze the posts of others and suggest helpful links. Recently, the bot got into some hot water with the community when it posted some links that, while not inherently offensive, were rather offensive in the context of the actual conversation.

To what extent is it reasonable to hold the writers of a bot accountable when their bot engages in insensitive behavior?

I’m conflicted on this. On the one hand, bots are imperfect and can’t be trained to handle every social scenario with perfect tact and etiquette, but giving bot writers a quick cop out ("My bot isn’t programmed to understand trigger warnings, so when it saw ‘Trigger warning: soldiers’, it had no idea that the OP didn’t want a link to a site where they could find their local army recruitment center. I shouldn’t be penalized!") also seems unreasonable, as each of us are ultimately expected to reap what we sow and one generally can’t "un-offend" someone who has already been offended, regardless of whether one intended to offend or had fully apprised oneself of the social context in which one was posting.

Is it reasonable to be extra-lenient on bot writers whose bots post inappropriate content (giving them more slack than an average "live" user who negligently, carelessly, or ignorantly violates rules or posts content likely to offend), or is it better to judge on content alone, requiring bot writers to either fix their bots or take them off-line under penalty of banning? For example, if the general practice of the moderators is to issue a one-month suspension for ignoring a trigger warning, is it fair to issue such a suspension to a bot writer who fails to add an adequate trigger warning detection script or is a different approach better?

One Answer

A bot user is designed to dynamically perform actions without the creator or a user.

In this case, you shouldn't hold the bot owner accountable for something a dynamic system generated, that he in no way meant to be generated.

Ultimately, you cant expect code to do everything perfectly, especially when it comes to linking outside content. "Punishing" the bot will do nothing, it doesn't learn from being "abused" and doesn't "feel" anything. The only choice is rewrite logic to improve its capabilities and recognition.

There is no need to discipline anyone over this, its a simple mistake.

However, if you really want to "punish" someone, just turn off the bot, or ask the creator to develop better logic for your chatroom if you have easily-offended users. I'm not trying to be rude, but this sounds overheated, and I don't think any "punishment" is necessary.

In my opinion, I feel it would be best if you asked the bot owner to apologize to the offended user(s) and ask if he/she could write different logic.

I will mention this again, when dynamically generating content (like links to outside content based on a search), you should expect for outside content to be unmoderated, at all times. This is why not many bots have such features, because you cannot control what will be created, and the bot could send inappropriate or offensive links because its simply finding content based off of the search terms.

Furthermore, please do not blame content recognition on a bot owner. Developing such a script isn't very easy, as all the bot can do is real URL text and similar meta info for content that is being linked, it cant use some magical code to recognize something inappropriate in an image. Such artificial intelligences that CAN read image data do exist, but are mostly proprietary AI software used by big corporations to moderate their enormous platforms.

This answer really is mostly biased, but Ive included some good points you should read, and give me your opinion if you disagree or are conflicted on something Ive mentioned.

Answered by Mister SirCode on November 27, 2020

Add your own answers!

Ask a Question

Get help from others!

© 2024 All rights reserved. Sites we Love: PCI Database, UKBizDB, Menu Kuliner, Sharing RPP