|
E: Apparently when I posted this I committed the cardinal goonsin of not reading the OP.  My quid pro quo review is in the next post, over the page.OK, here goes. This is a little something I'm eventually planning on submitting to Nature for their Futures section. The brief is 850-950 words words of hard sci fi. However, since I wrote this based somewhat on my own experiences in data science, which even I find a little dry at times, I'm particularly interested in knowing whether a) the jargon-o-metre is in the right place between understandability and hard sciencey-ness, b) the story is compelling, and c) the character is relatable to. All other comments are welcome. Deathwatch (944 words) Daniella drove down the cracked, too-wide street in the barren Detroit neighbourhood housing the United Credit data centre. Some days, she wondered why , but the response came quickly: the enormous salary premium. The money UC saved by using a disused telephone exchange left them with more than enough to afford it, and protecting profits through cost-savings was the whole point, wasn't it? Besides, the work was too interesting, the data too big, for her to possibly pass it up. In fact, she suspected that part of UC's decision to base their main data mining centre in a run-down corner nobody wanted to visit was precisely to keep it low key. She had the full purchase histories, credit records, demographics (actual and inferred) and online activity logs on every one of UC's nearly fifty million customers at her fingertips, and already she had worked magic with it. She reminisced over her achievements as she passed the discrete but formidable perimeter security and into the Faraday-caged office area, her phone bleating plaintively about the lack of signal. Marriage had been her greatest success to date. With the right combination of features -- certain purchases, subtle changes in credit rating indicative of emotional distress, income for striation -- all normalised to county-level census data and trained with the right classifier, she could predict when a customer was about to marry (with cross-validated PPV of 0.98), or divorce (PPV=0.97). Working with the production team, she had optimised the classifier for speed, validated on a follow-up cohort, and rolled it out as a secure internal service to account managers country-wide. She'd had some doubts about the sales execs' idea of increasing credit limits to cover wedding or divorce costs, but UC did also offer preferential rates on florists, catering, honeymoon vacation packages, divorce lawyers and psychotherapy. Besides, the intellectual challenge of the problem was what really moved her. The latest project, now that was a challenge: suicide prediction, or "deathwatch", as they'd begun to call it. She had no idea where UC had dug up the training data -- data sharing agreements with life insurance companies, probably -- but that had been the least of her problems. She'd had to cast her net wide for inputs, and feature selection had been tricky. Still, the right types of credit card payments, web browsing trails winding through suicide prevention websites, and social media posts both from the imminently deceased and their "friends" could be most revealing when sifted from the noise. She'd even run literature miners over the suicide research corpus. And finally, it had worked. She had a developed a rock-solid, tightly bounded score for six-month suicide risk. For a while, after UC had rolled it out and the first lives had been saved by trained (credit) counsellors, she had begun to feel like she was changing the world. But this morning, as she sat at her plush, three-monitor workstation, something had been bothering her. She had been running diagnostics on the deathwatch predictions since the interventions, and the positive effect was quite visible. Yet, quite a few of the predicted suicides seemed to happen anyway, and she was determined to find the pattern. As she set up the analysis on the compute cluster, she idly flipped through some of the other data scientists' code on the network drive, her eye stopping on a folder titled "Suicide Cost-Benefit Analysis". Determining exactly what the main script was doing would take some time, but the main function took in a personal identifier, and spat out a score. She had a hunch, and set up a quick run over a subsample of the suicide predictees. The score gave an almost perfect stratification. She stormed into her manager's office. "You're just letting them die!" "Daniella, Daniella, " "I found the cost benefit script. I know. When they aren't valuable enough to UC, they don't get an intervention." "Come now, the counselling costs money, and we have an obligation to our shareholders to turn a profit. We have to know when the cost of the counselling exceeds the predicted future lifetime profit from the customer. And it's not like we're heartless; we send them a letter suggesting they get counselling. You should know ..." Her mind numbing as the impenetrable miasma of business jargon enfolded her, she nodded, meakly, in compliance. Eventually she shuffled back to her desk in a daze, and too drained for real work, opened her inbox to an email from Mom which turned her blood cold. Dad had received a strange letter from the credit card company, suggesting he look into counselling. Suddenly, she was alert again. The records were de-identified, but picking Dad out was easy enough. Her palms sweated as she began running him through the suicide risk predictor, adding a new refinement she'd been working on to predict time of suicide. Her clenched fingers slipping on the keys, she started the cost-benefit script in parallel. The prediction confirmed her fears -- his suicide risk was through the roof. The cost-benefit analysis hardly surprised either. Dad had always been careful with his money. Eyes transfixed to the screen as the time predictor churned away, Daniella's thoughts raced. Should she run to the smoking balcony so she could call home? The sudden flicker of the script completing decided her, as it predicted a single day. That day. She floated to the balcony, her damp fingers slipping on the touchscreen as she dialled home. Each ring drew the knot in her stomach tighter, and Mom's frenzied voice rang out like a gong. "Daniella? Daniella! It's horrible.... your father ... shotgun ... outside ... shed ..." "He's dead, Daniella! Dead!" Lead out in cuffs fucked around with this message at 04:20 on Feb 1, 2014 |
#
¿
Jan 30, 2014 07:23
|
|
|
|
| # ¿ May 14, 2024 16:21 |
|
|
Nettle Soup posted:Can I get some basic advice on this? It was a 3 random words ~500 word challenge to myself. It's grown to about 700 now so I'm going to stop editing it and just post. I'm heartened the thought that it can't possibly be as incomprehensible as that mess of words above. Overall, I like it, though I'm not sure all goons will. The nostalgia and whimsy are sufficiently cloying that I think you can scrape by with the relative lack of drama. There are some clarity problems, though, especially with the ending. You could make us care a lot more, too. E: corrected minor bolding error. Lead out in cuffs fucked around with this message at 04:15 on Feb 1, 2014 |
|
#
¿
Feb 1, 2014 04:13
|
|
|
Nettle Soup posted:Thank you! I realise it has some major issue, but if nobody points them out then I'll never improve! I think the main problem is, it started out as a guy giving his girlfriend a necklace he couldn't possibly afford and her thinking he'd stolen it, and then suddenly the woman shoved her way in and I couldn't think of a better opening. I should have just erased it and started fresh. You're welcome! I hope it didn't come off as too negative. As I said at the end, I do kinda like it, it just needs some work for consistency and clarity. Maybe work on explaining the narrator's relationship to Joan, and think of some kind of better ending -- either work the locket angle (it's your main theme, no?) with the new neighbour, or else something happening with Joan's family. Or hell, have something happen with the locket. Have the narrator arrested, or be currently in jail. Anything but her taking some kind of spiteful vengeance on her own family we've never heard about until the last sentence. Nettle Soup posted:She never found the headstone because the family didn't contact her, and they didn't have one made anyway, just kept her ashes on the bookcase. Hmmm ... but wouldn't she have gone to the funeral? They're generally public, and if she cared so much, she would've found the details and gone, even if not invited personally.
|
|
#
¿
Feb 1, 2014 23:30
|
|
|
Their writing is clunky in part because they're making a lot of Russian ESL grammatical errors and weird turns of phrase. It's also that they're telling rather than showing. Blah blah is a military unit. Blah blah is a legendary creature. Show, don't tell. Don't get too flowery, but create a little tension. [Greekos Commanderos] wiped the snow from his visor and gazed down the slopes of the West Oros to the ruins of [blah blah] below. His most stalwart scout had returned, shaking and pale. Any other man he would not have believed, for the being he described was a myth, something to scare children at night. But if the legends were true, and it should reach its lair at [blah blah]... His veins turned to ice as he thought of his mother, knitting by the fire back in [Greekia citia], unaware of the doom hanging over them all. That could probably be more succinct, and I'm making up details because I don't know your world, but it's a hell of a lot more interesting than the way you have things written right now. Also note how much information is embedded in the descriptions. That first sentence tells us he's a soldier (helmet/visor), that it's winter (snow), that the Oros is a mountain range (slopes), and that the blah blah lair place is a ruin. And that was without a single "is a". Something you also need to think about is your character building. Is the commander an important character? For the first half of your piece he's the centre of focus. If gryphon lady is your sole main character, and the commander guy is just a mook, you should probably trim down the fluff at the beginning and get to her as soon as possible.
|
|
#
¿
Nov 13, 2014 19:27
|
|
|
Erik Shawn-Bohner posted:3) If you want a critique, post one first. 1:1 ratio. Just a little reminder that while Thunderdome regulars have been incredibly magnanimous in giving out crits for free, this thread is meant to be quid pro quo. You should still be posting a critique of another story at the very least as an exercise in improving your own writing.
|
|
#
¿
Feb 6, 2015 18:09
|
|
|
|
| # ¿ May 14, 2024 16:21 |
|
|
Just a quick one, since it stood out:TheForgotton posted:“Snap out of it, Marsh. That's how they get you.” Clara's voice. Marshall stumbled but regained his rhythm. Another couple of seconds, and he might have stopped dancing entirely. He heard her too clearly over the dissonant waltz but felt no breath touch his face. He whispered his thanks into the crook of her neck and tried to focus his thoughts away from his slavering belly. Slavering means drooling. Unless you're doing body horror, bellies do not drool.
|
|
#
¿
Jun 8, 2015 19:47
|
|